Report On Possibilities For A Photograph Database by Adam Engst' Consultant 901 Dryden Road #88 Ithaca. NY 14850 Introduction As Gould Colman explained it to me, the transfer of the bibliographic information on the archival photo- graph collection to a computer database is currently open to a large number of possibilities. While there are advantages and disadvantages to programs on a number of machines, there are no external forces currently requiring a certain solution. I was retained to research the possibilities and present my findings, either recom- mending hardware and software combinations to test or recommending that the Archives wait several years before repeating this process. 1 have gone through several steps to come up with this report First, I tried to determine precisely the needs and desires of the Archives. Second, I researched the software possibilities on three different hardware plat- forms which are available to the Archives, the Macintosh series, the IBM PC line, and a general category of mainframe. Third and finally, I weighed the advantages and disadvantages of various combinations, adding in my knowledge about the Cornell community, the state of the software industry, and the history of some companies in particular. Most of the information below comes from my notes on telephone conversations held with represen- tatives of the various companies. Questions Since the Archives is not locked into using any specific program or computer, I was left to figure out what sort of a system would best fit the needs of the department As I see it, there are some requirements placed on any system by the size of the data, the nature of the data, the use to which the data is put, and the cost of the hardware, software, and programming time. Size and speed Gould told me that the Archives currently has between 50,000 and 100,000 photographs. Assuming one record in the database for each photograph, the system must be able to handle 100,000 records with decent searching speed. This was the first question I asked of the various database companies. Their replies must be taken at face value though, since the only way to really test the each system is to put 100,000 representa- tive records into each and do some searches. Keywords Since a small number of photographs are the end result of any search through the database, the system must be able to handle a relatively large number of keywords, or else researchers will have difficulty narrow- ing down their searches. A number of databases require programming convolutions to be able to deal with a field containing an unknown, but potentially large, number of keywords. Such a hmitation does not rule out a database, it simply downgrades it in terms of ease of setup and programming. In addition, selecting the keywords is an extremely important task which must be thought out carefully. Graphical information Bibliographic information is useful for providing a brief description of each photograph and locating it within a collection, but for a researcher who is trying to find a certain photograph, possibly out of hundreds of similar ones, bibliographic information will not allow that researcher to select a certain photograph with surety. A graphical method of describing the photographs would decrease the amount of time it would take a researcher to find the right photograph for the use he or she has in mind. The main possibility is displaying images on a videodisc. Few of the database systems can directly control a videodisc player. There are serious cost drawbacks to displaying visual information though, so inability to control a videodisc does not disqualify a database. Initially, scanning the images and storing them on a CD-ROM would seem to be feasible because each CD- ROM can hold 600 megabytes of information. Unfortu- nately, CD-ROM is not feasible because a scanned photograph has an average file size of 300K, which, when multiplied by 100,000 photographs, would force you to use close to 50 CD-ROMs holding 600 megabytes each. In comparison, a single videodisc can hold 108,000 images. Costs The cost of the software are minimal in compari- son to the costs of transferring the images to videodisc, although the purchase of expensive mastering equipment can reduce the overall costs. Another cost which cannot 46 lASSIST Quarleriy be ignored is the cost of programming and setup with whatever software is decided on. As a result, ease of programming does play a financial role in the final decision as well. So the questions that I asked each database company were as follows: •Can your program handle 100,000 records with a fast search speed for a single record, say under 10 seconds as a worst case scenario? •Can your program handle unlimited length text Fields in an index (to retain searching speed) or is there a simple way around the program's inability to do so? •Is there any way for your program to access images stored on a standard videodisc player? •How hard would it be to set up your program with a simple interface for researchers who may be inexperienced with computers? I also tried to get a feel for each company—how easy they would be to work with, how much help they would be if we needed any technical support, and whether or not they would still be in business in several years. These are intangibles, but potentially useful pieces of information. A note before I get into the details. I've tried to write this so no technical knowledge is required to understand it I'm sure that in some places I have failed because there is simply no other way to talk about certain features and actions of computers. In those places, I've included a footnote or tried to explain the term I use within the text. If at any time, you are confused reading this, please call me, and I will attempt to clear up the source of the confusion. Hardware The companies with whom I spoke have database programs that run on the Macintosh series of microcom- puters, the IBM PC line of microcomputers, and (in the cases of Oracle and NOTIS) almost all minicomputers and mainframes. The Archives currently has several IBM PCs and clones and will be getdng several Macin- tosh SE/30s shortly. In addition, I gather that the depart- ment has access to the mainframe resources of the library and the university. So existing hardware does not bias the decision. With a few exceptions, all of the software pack- ages I researched can handle the large size of the data- base without a loss in searching speed. Obviously, the minimum (and preferred) hardware configuration in each case does vary slightly, although some packages run fine on less powerful machines, which is a bonus since it will reduce the costs. In general, and like all generalizations this one is not to be trusted completely, the Macintosh will be the easiest to set up and for both researchers and staff members to use. IBM PC clones have the advantage of being in the majority, although powerful systems are not really much cheaper than Macintosh systems. Micro- computers have the advantage (and disadvantage) of local control—if something goes wrong with the com- puter you can have it fixed quickly if necessary, whereas you must wait for another department to respond to your problem with a mainframe. On the other hand, if some- thing with a microcomputer fails, you must deal with it, unlike with a mainframe, which will have a staff to deal with problems. Mainframes often suffer from poor interfaces as well, although there are ways of avoiding the poor interfaces. Using a Mac and HyperCard with a videodisc is simple, while using a PC with a videodisc requires a special device driver, which is a small program which allows the computer to control the videodisc. Such programs are available, often from the videodisc maker, and there are also programmers who could write a custom device driver if necessary. General hardware conclusions Based on my experiences with the various types of computers and my knowledge of the Cornell user community, I recommend using a Macintosh. Macs are predominantly easier to work with in the setup phase, and they are far easier for inexperienced users to work with. Since the entire point of this project is to provide easy access to information, I think that the interface is one of the most important parts of the system, and better interfaces can be created on the Macintosh. In any case, my research covers all three platforms, and I hope that my recommendation of a hardware platform is bom out by the software possibilities on the Macintosh. In addition, the Macintosh database companies were far more knowledgeable about controlling videodiscs, which is why I often have more information on the Macintosh databases. Macintosh Software Company: IstDesk Systems Program: IstTeam Hardware: Mac Price: $795 Their powerful relational database, called IstTeam, is compatible with HyperCard and can store up to 255 characters in each field. Offhand, 255 characters doesn't sound like it would necessarily be enough for our keywords, but Spring 1991 47 perhaps their literature will shed more light on the subject. Otherwise, IstTeam is certainly a possibility because it should be fast enough and can control a videodisc through HyperCard, although it is much less well-known than either 4th Dimension or Omnis 5. Company: ACIUS Program: 4th Dimension Hardware: Mac Price: $695 4D can handle 100,000 records with no problems, but it would have trouble with indexing. Without indexing, the search speed slows tremendously, but 4D cannot index its unlimited length text fields, which we would use for holding keywords. So setting up the keywords would be a litUe tricky in 4D. There are a number of different ways around for this problem, but they would require a bit more work programming. In the first French version, there was some kind of external command which could control a videodisc, although they may not still exist in the current version. There is a demo database, called Minifans, which we could look at if 4D turned out to be a likely candidate. Against 4D, I've heard that it is one of the slower databases for the Mac, which is a problem for this project A test of its speed would definitely be needed before I could recommend it any farther. ACIUS is one of the major database companies for the Mac, but I have been unable to get through to them at all, which may indicate mediocre customer support While this is not a complete argument against using 4D, it doesn't bode well for future support needs. I can't really recommend them unless I can get through to talk to them. Company: BIyth Software Program: Omnis 5 Hardware: Mac and PC Price: $695 Omnis 5 from Blyth Software certainly has the power to deal with 100,000 records, and it can be extended to do even more such as control a videodisc, although the representative didn't think such an external command had been written so far. Alternately, either Blyth could do it for us fw free if it was small and fairly easy or an independent programmer might be willing to write such a thing for a fee. Omnis can index variable length text fields, so it would have no problem with a field containing a variable number of keywords. If the software to control a videodisc was difficult to write or acquire in other ways for Omnis, it can work with HyperCard so that HyperCard uses the Omnis database while acting as a front-end. However, Omnis can also create simple interfaces easily, so it should not be necessary to link the two together on that account. Omnis's language is supposedly English-like and easier than most database programming languages. An advantage of Omnis over any of the HyperCard extensions is that Omnis is a full-fiedged database, and as such, can generate reports and display multiple windows, which would be good for displaying a number of records which met the search criteria. Omnis needs a minimum of 1 megabyte of memory and is happier with a fast machine and more memory. Overall, I was quite impressed with the possibilities of using Omnis 5, since it seems to meet all the requirements and be fairly easy to work with in addition. The representatives have been extremely knowledgeable and responsive, unlike some of the other companies, such as ACIUS. Company: Fox Software Program: FoxBase Plus and FoxBase/Mac Hardware: PC/Mac Price: $395/$495 FoxBase can have up to 254 characters in text fields, and can search on unhmited length text fields, but they aren't indexed which slows the search. However, Fox claims that FoxBase can handle up to 1 billion records, and that it is the fastest of all the Mac databases by a great deal (some 30 times faster than 4D), although some of the PC databases come close in speed. Reportedly, a new version of FoxBase/Mac can use HyperCard's external commands (such as the ones to control a videodisc player) directly, which would be a major point in its favor. FoxBase is generally accepted to be better than DBase Ill-t- on the PC and the Mac version is cwrespondingly good, if not better since the Mac version can handle unlimited length text fields. FoxBase cannot control a videodisc, although it could work with a CD-ROM. Despite FoxBase's speed and file compatibility between machines, I think it is somewhat too Umited in this situation because of its inabiUty to index unlimited length text fields and its inabihty to control a videodisc. In addition, if it crashes for any reason, it will often corrupt the entire database rather than just losing the last record entered. This is a serious problem because you can never predict crashes. Company: Odesta Corp. Program: Double Helix 11 Hardware: Mac ASSIST Quarterly Price: $395 Double Helix cannot link to HyperCard and cannot control a videodisc, although there is a version that runs on Vax mainframes (for about $5000) which would help the speed and storage problems. Despite the fact that Double Helix can handle 100,000 records and has a simple method of programming, I doubt that this program is a real answer. Double Helix simply doesn't have enough to recommend it over any of the other major databases except its idiosyncratic programming environment, which may be easier than most HyperCard extensions All of the following jjroducts require HyperCard, or one of two HyperCard clones, SuperCard or Plus, which provide the same basic features as HyperCard but with significant extensions. If a HyperCard system is decided on, it would be well worth the time to investigate creating the database in SuperCard or Plus rather than in HyperCard itself. The various extensions hsted below may or may not work with SuperCard or Plus, although there is a good chance that they will. The areas in which SuperCard and Plus go beyond the capabilities of HyperCard include reporting, graphics, multiple win- dows, and color. Whether or not these features are worth moving away from HyperCard is another question entirely, and one that need only be asked if the Archives decides to go with HyperCard rather than one of the full- fledged databases. Company: Answer Software Program: HyBase (under HyperCard) Hardware: Mac Price: $150 Size and speed are not problems, since HyBase can handle up to 2 billion records and can usually find a single one in about 5 seconds. All fields are unlimited in size, or at least very large. The company claimed that it is not difficult but that some programming experience is helpful. Answer Software could set up the database for us if necessary. However, Gregory Crane at Harvard said that he used HyBase on Project Perseus for a while and found it very difficult to work with. Dealing with the people at Answer Software was rather difficult and based on Gregory Crane's advice, I don't think that HyBase is a good possibility. It suffers from difficult set up, which is unnecessary for this project Company: Discovery Systems Program: HyperSearch (under HyperCard) Hardware: Mac Price: $99 I have not yet received any information from Discovery Systems regarding their HyperSearch package, so I cannot make any specific statements for or against it. However, Library of Congress is using it in their American Memwy project, and they seemed pleased with its speed and ease of use. Company: KnowledgeSet Corp. Program: HyperKRS (under HyperCard) Hardware: Mac Price: $195 HyperKRS works completely within HyperCard so it would be simple to design the database. Nothing else need be done in terms of setup except for generating the index, which is fairly slow, but only needs to be done once. A Mac Plus is all that is required for searching. HyperKRS was designed for CD-ROM, which accounts for its speed. I tested the demo software they sent me and I wasn't remarkably impressed. I had trouble finding anything, mostly because I was unfamiliar with the information for which I was searching. The speed was good but not great, but my Macintosh is not that fast, which is certainly an issue with this program. On the whole, HyperKRS sounds like it may be the simplest of all the HyperCard extensions to set up initially. After that, I have no real numbers to compare its speed with HyperHIT or Xearch. Company: NovaSoft Engineering Group Program: GridFile (under HyperCard) Hardware: Mac Price: $195 NovaSoft has a sample application called ClipFile for GridFile which is being used right now to access pictures on clip art CD-ROMs. We would have to do the indexing and setup ourselves, which would be difficult without the aid of a relational database expert. GridFile is extremely fast, though, and is able to search any database for a unique record in 3 disk reads (certainly under 1 second). The cons of GridFile include the fact that it requires 2 megabytes of memory and a fast hard disk; it does not provide as good data packing as some other databases, which makes the file larger, it slows down on smaller databases in comparison to the others; it doesn't support split files over two or more hard disks; and it would be hard to set up. The pros of GridFile are that it is blindingly fast (faster even than some mainframe databases) and that it uses HyperCard as a front-end, which can then control a videodisc. Spring 1991 On the whole, I think GridFile is very powerful, but possibly too difficult to work with. There are other programs which provide similar speeds, but are easier to work with and require less hardware. Company: SortStream International Program: HyperHIT (under HyperCard) Hardware: Mac Price: $195 Steve Hannaford, the technical support representative for HyperHIT said that HyperHIT has extremely fast searching speed (~1 sec) and can handle unlimited length fields with no problems. The information does not need to be textual—it could be pictures or sounds. The setup is not trivial but not that hard, and the HyperHIT system is entirely contained in external commands that work within HyperCard. Steve didn't think it would be hard for someone without database training to use. The data file is external to the HyperCard stack which would control the videodisc and thus requires only a small amount of space. Another advantage to the external data file is that there could be a number of different interfaces since the HyperCard stack does not have the data embedded in it. Search time is usually under 1 second with a Mac Plus, and it would drop with a faster machine and hard disk. There is httle speed degradation when the file size increases (3 hundredths of a second when going from 1000 records to 10,000 records). In fact, the videodisc might be the bottleneck, depending on how fast it can find each frame. Steve Hannaford was very helpful and said that he wasn't getting many calls as the technical support person for HyperHIT, which could mean that people aren't having any problems worth calling about. The advantages of HyperHIT are that it is extremely fast despite what sort of machine it runs on, it supposedly isn't difficult to set up (although Gregory Crane will be testing it for Project Perseus soon and will have an opinion on its ease of use), and it will allow simple interfaces and videodisc access through HyperCard. Overall, HyperHIT sounds like a good possibiUty. Company: The Voyager Company Program: VideoStacks Hardware: Mac Price: $99.95 VideoStacks is a set of external commands to control a videodisc for a number of different videodisc players. There is a possibility that some of the external commands would be available from Apple free of charge or they might be distributed with the videodisc itself. Some sort of videodisc drivers will be necessary. Company: Xiphias Program: Xearch Hardware: Mac Price: $??? Xearch is an external command for searching in HyperCard which Xiphias uses in their CD-ROM- based product. Time Line of History. It sounds like it would be fast enough and they do have a licensing agreement, although I don't yet have the details. Initially Xearch sounds like it could be quite useful, although 1 don't have a sense of how easy or fast it is in comparison to HyperHIT or HyperKRS. General HyperCard software conclusions I think that all of the various packages mentioned above will probably provide HyperCard with the search- ing speed necessary to use the database. The main distinction then, lies in the ease with which each is set up. HyperKRS and Xearch are probably the easiest, with HyperHIT, HyBase, and GridFile lining up in increasing order of difficulty. More specific research and testing would need to be done to determine speed and ease of use in order to choose between the various extensions. HyperHIT may be the best compromise between speed and difficulty. General Macintosh software conclusions I am of two minds in this category. I think that HyperCard is a wonderful program (not to mention the fact that it is free with all Macs), and it will become integrated into the Macintosh hardware and system software in the next few years, making it even stronger and faster. On the other hand, it really is not a database and does not provide the features that a full-fledged database provides, such as reporting and fast searching. It may be necessary to use HyperCard in some fashion to facilitate access to a videodisc, which lends strength the cases of those database products that can link to Hyper- Card, such as 4D, Omnis 5, and IstTeam. On the other hand, the structure of the proposed database is very simple and does not really require the full power of a relational database. In the final consideration, I think I would currently recommend Omnis 5 because of its power, flexibility, and ability to link to HyperCard, not to mention the quality of the customer support, with which I was very pleased. PC Software Company: Ashton-Tate Program: DBase IV/Dbase Mac lASSIST Quarteriy Hardware: PC/Mac Price: $795/$49S Neither DBase IV nor Dbase Mac have any internal way of controlling a videodisc, and the representative didn't know of any external ways either, although he thought one might be possible. In addition, neither can index on variable length text fields, which would slow them down a great deal for this purpose. Add these problems to the fact that Ashton-Tate is undergoing major problems as a company and has publicly announced that they will not be upgrading Dbase Mac at all, and you get a company to stay away from. Company: Borland International Program: Paradox/Reflex Plus Hardware: IBM PC/Mac Price: $725/$279 The Borland representative didn't think that either Paradox, the more powerful PC program, or Reflex Plus, a decent Macintosh database, could control a videodisc. It took several phone calls and some time on hold to get that much information, so I didn't pursue it farther. However, Tim at Turquoise FilmA'ideo Productions (one of the mastering services) said that he was thinking about re-writing his custom database in Paradox because it was fast and fairly easy to work with. He also said that Paradox runs on a number of machines and is probably file compatible with Reflex. As a result, Paradox sounds like the best of the PC databases that I've looked into. Using Paradox would require some additional device driver to control the videodisc, but such a program might be available from a number of sources, including Turquoise Productions. Paradox won most of the speed tests I saw in the course of my research, so I would recommend it over the other PC databases based on what I currently know. Company: DataEase International Program: DataEase Hardware: IBM PC Price: $700 DataEase cannot control a videodisc, although it supposedly can interface with three scanners for using pictures. However the storage of those scanned images would be ridiculous and dealing with graphics on the PC is more difficult than on the Mac. DataEase does have long text fields which can be searched, although the representative wasn't sure about whether or not they were indexed, which is a major concern. I have a demo disk from them which may answer the indexing question, although I see nothing special about DataEase otherwise. Company: Image Concepts Program: C-Quest Hardware: PC or Unix mainframe Price: $6000 or $25000 C-Quest is a proprietary system for storing photographic information and controlling a videodisc. It has been around for several years, but doesn't seem to have a devoted following. Clif Nickerson of Image Concepts was somewhat helpful, although his system is designed more for a stock photograph collection than a histOTical research collection. The main evidence of this is the way it uses synonyms of keywords, a method which allows the user to search on "Stream" and get "Brook" and "River" and "Run" and "Creek". Unfortunately this is not nearly as useful with proper names of people and places, since they tend to be specific. The only use I can think of it is use modifiers, so you could have Frank Rhodes walking, talking, shaking hands, or making a speech, and search on the action involved. I don't know if that is too much trouble to set up and key in or not C-Quest runs under Unix mainframes as well as PC clones. Under the Unix system, C-Quest can display 18 pictures at once; on the PC it can only display one at a time. Its speed is dependent on the number of subjects used in the search, but Clif said something about speeds of under 1 second, which he said was faster than the mainframe database Ingres (and he thought than Oracle). The C-Quest interface is menu-driven and not particularly good. It does not have a simple interface for researchers to use, although one is being proposed. Image Concepts will change, add, or remove fields from the menus for a nominal fee, which is not as good as setting it up oneself. C-Quest is not cheap, by any means, at $6000 for the PC version of the software and $250(X) for the Unix version. Clif said that the easiest way of getting images on disc were to buy a video camera and a writable videodisc, at which point you could do it all in- house. I suspect that quahty wouldn't be as good, although there is no way to know without trying. He recommended making a 35mm film image in case the resolution of the monitors increased enough to make it worthwhile to re-master a videodisc. C-Quest has an impressive list of clients, although 1 suspect that is from being the only game in town for 4 years, since no one else does this on the PC at all. Evidently, the Library of Congress system is slower than C-Quest, although that doesn't really Spring 1991 mean much without more details. C-Quest can control videodiscs from a number of companies, such as Sony, Pioneer, Philips, and Panasonic. Overall, I find their system to be somewhat clumsy, expensive, and not really suited to the needs of the Archives. The Archives photographs have specific subjects without synonyms and need a very simple interface for researchers. While $6000 is not truly expensive in relation to the cost of mastering the videodisc, I think it is quite a bit more than you would pay for any other system. It might require less setup initially, although it is still a generic program that would require some customization. I cannot recommend it, especially since I heard from another consultant that the version of C-Quest at the United Nations was actually quite slow. Company: Microrim Program: Rbase for DOS Hardware: IBM PC Price: $725 Rbase will handle an unlimited number of records but the representative didn't know if it could handle an unlimited length text field. I didn't want to hold any longer to find out if it might be able to control a videodisc since the representative didn't think so. I see no reason to specifically recommend Rbase. Company: Symantec Program: Q&A Hardware: IBM PC Price: $??? Q&A does not have variable length text fields, but it can have large ones which are indexed so that the search speed doesn't suffer. The speed isn't good, though, at 15-20 seconds average, partly because Q&A is not a high-powered relational database. There is no videodisc access, although the representative thought that an external program might work. Considering the.speed problem, I can't recommend looking any further at Q&A. General PC software conclusions I think most of the major PC database programs will handle the textual part of the database without trouble. However, it seems as though it will be more difficult to link the textual information in a PC database to the frame numbers of a videodisc. These device drivers do not seem to be readily available or supported by the database companies. For instance, the representa- tive of Ashton-Tate knew nothing about linking to a videodisc, yet supposedly the Library of Medicine is using DBase Ill-t-. However, a device driver might come with the videodisc player. I also feel that it will be more difficult within these PC programs to create a foolproof interface for researchers who are inexperienced with computers. I missed at least two major databases for the PC, Revelation and Nutshell, because I was unable to find phone numbers for them. However, I am not particularly wcxried that they are the perfect database because none of the other PC database companies had much of an idea what a videodisc even was, much less if their program could control it. The PC database compa- nies were also much harder to reach on the telephone and much less willing to talk. If someone else turns out to be using Revelation or Nutshell, it would be worth checking them out. Otherwise, I think Paradox will be the best on the PC side. Mainframe Software Company: Oracle Program: Oracle (runs under a HyperCard front end on the Mac) Hardware: Mac/PC/minicomputers/mainframes Price: variable depending on version—from $299 to $1299 Oracle for the Macintosh is a port of the most popular database program in the world. It retains complete compatibility with all other Oracle databases on all other machines, which is a plus if this data will be shared with other people. In addition, a Macintosh running the HyperCard front-end to Oracle can use any Oracle database on any machine. Because HyperCard is the front end to the actual database, Oracle can control a videodisc through HyperCard. Should the Archives wish, they could probably find a mainframe or minicomputer on which they could use Oracle. Oracle's advantages are speed, portability, and ease of use with HyperCard, although it might be a bit more expensive than the Archives would want initially. There is a Developer's Version for the Mac for $299, which would allow us to test its capabihties (with the only limitation being that this version cannot link to other Oracle databases on other machines). The main disadvantages to Oracle are that it is potentially more expensive (although the Macintosh version is quite cheap) than other databases, and that it may simply be too complicated for the relatively simple database information we have. I gather that setting up a database in Oracle is not all that easy. In addition, I've heard that Oracle for the Macintosh is not that fast and occasionally does strange things to data files. Oracle as a company is excellent, with toll free support and guaranteed stability. They are the 52 ASSIST Quarterly largest database company in the world and the third largest software company in the world. Company: Northwestern University Program: NOTIS Hardware: IBM mainframe Price: free NOTTS has a number of advantages, although it also spOTts major several disadvantages. NOTTS is currently installed and running in the library, so there are no added software or hardware costs to the system other than a Macintosh and videodisc from which people can search in the Archives. Tt is relatively fast and can certainly handle another 100,000 records in its database. NOTTS has the advantage of being accessible from anywhere on campus, but researchers may not use it unless they can also see the videodisc images because it is difficult to search for photographs based solely on bibliographic information. It has been in use at Cornell for some time now, so many people are familiar with its interface, although its interface is also one of its main disadvantages. Searching and moving between the various results of a search in NOTTS is difficult and completely not intuitive. Its other main disadvantages include the fact that it would be very difficult to link it to a videodisc, if it is possible at all, and the problem of portability of data since NOTIS does not have the abiUty to export its information to another program, something which all of the microcomputer databases can do and which is very important for future expansions or modifications. It might be possible to sidestep NOTTS's poOT interface with a HyperCard interface currently being worked on at Mann Library. In addition, there is a commercial product that will be available soon from Texas A&M and Apple, called MacNOTIS, which also provides a better interface to NOTIS. Because both of these fM-oducts use HyperCard, it is theoretically possible to have the HyperCard interface control a videodisc while using the information from the NOTTS database. Howard Curtis of Mann Library thought that this was possible, although extremely clumsy and prone to break whenever either NOTTS or HyperCard changed much. Other people thought that it would be an unworkable situation even if it was theoretically possible. Howard also said that it might be possible, though difficult, to program HyperCard to download records from NOTTS to the Mac, which would allow the records to be used by microcomputer databases. NOTTS is an easy solution because it requires no new hardware or software, but putting the records into NOTTS removes them from a certain level of accessibility. Tt would be difficult and clumsy to attach a videodisc to a Macintosh running one of the HyperCard interfaces, if it is indeed possible at all. More research would need to be done to determine tJie reality of such a setup. Even worse, it would be hard to transfer those files to any microcomputer system. However, there are some ways of moving from microcomputer databases to a format which NOTTS can read, which points towards putting the records into a microcomputer database first, and then, if there is interest, transferring a copy to NOTTS. As much as NOTTS seems like the simplest solution, I don't feel comfalable recommending it given the possibility for videodisc access and the inaccessibility of the data once it is in NOTTS. I reaUze that NOTTS data can be shared by other mainframe cataloguing databases, but they don't (on the whole) provide the kind of features that microcomputer databases do. Being able to move data between systems is important, and customized mainframe databases are a blockade to such a move. Videodisc Mastering Services Company: Image Premastering Services Program: Videodisc services Hardware: NA Price: variable Image I'remastering Services claims they are known for having the highest image quality for still frame transfers. Some of their main clients have been the United Nations, the Library of Congress, the Mayo CUnic, and the American College of Radiology. They can handle absolutely any original—for the Library of Congress they laid down 30,000 glass plate negatives without cracking any. Of course, slides are the cheapest method, and run anywhere from 550 to $1.35 per slide. Other original media are correspondingly more expensive, although presumably the cost goes down with quantity. In addition, there are some basic initial costs which cannot be avoided. These costs total $3 1 50, although that is minor compared to the cost of transferring 100,000 photos to the disc. $2000 for the master disc $500 fw a check disc $150 fw the videotape $150 fw a duplicate/backup, kept at their site $350 as a basic setup fee They claim that Stokes is mainly a slide copy service and makes a 35 mm film negative, which is a second-generation picture of the original. If the Spring 1991 image is from a print, then the videodisc image is third-generation picture and suffers correspondingly in quality. Stokes does color- correction, so the colors may be bright, but they are likely to be inaccurate. Stokes is also generally cheaper because everything is automated in their process. Image, on the other hand, is specifically dedicated to mastering videodiscs and they have patented technology for the process. They use a 2 foot lens over a 12 foot optical bench, which gives them two advantages. 1) The light comes in at a perfect 90° angle, which gives much better edge definition to the image. 2) They use an aerial image transfer, which somehow projects the image so that there is no film grain in the resulting videodisc image. It also allows them to easily perform custom sizing. The Image representative recommended the Mac and said that a military project used HyperCard with Oracle. He had heard something about 4D, but didn't know of anyone who was using it. He didn't recommend using the PC at all because the hardware is more expensive and is harder to set up the software to interface easily with the videodisc. Company: Stokes Mastering Services Program: Videodisc services Hardware: NA Price: variable I spoke with John Stokes and Jim Couch of Stokes Mastering Service. In regard to costs, Stokes estimated that a basic image transfer of positive images would be somewhere between $2 and $3 per image, although his estimate for complete costs (ie. in-house handling and database work) was closer to $4.50 per image. The work is done on- site and includes a person to come and do it with Stokes's somewhat specialized equipment There is not much difference between Stokes's doing the work and it being done in-house except for the fact that he claimed they had higher quality control, which is fairly likely. I suspect this is somewhat cheaper than Image Premastering's prices, although not by as much as I had originally thought They have a number of projects going on, the most notable of which is for the Library of Medicine, whose cost was about $2.40 per image. That price is slightly inaccurate because it was a test run in some ways and the library got two sets of negatives and slides for each of 70,000 images. The perscHi to talk to at the Library of Medicine is Lucy Kiester, phone number 301-496-5962. The Library of Medicine is using a PC with DBase III+ for their database. Stokes claimed that videodisc access was incredibly simple with any database and that you didn't need a custom driver, but he did admit that you had to write some software. The Library of Congress is also working with Stokes, which is curious since the representative at Image Premastering said that the Library of Congress was working with them. Perhaps there are two different departments? In any case, Stokes claimed that the Library of Congress is using some in-house computer system rather than an off-the-shelf software package. Stokes said something about how that was their policy. This is not necessarily true since the American Memory project is using a Macintosh and HyperCard. One advantage of Stokes's method is that you can get negatives of each image as well, which allows you to reduce handling of the original images by making additional negatives. The Library of Medicine uses these negatives for public access to avoid giving out their originals. Interestingly enough, Stokes said that more people are doing the imaging first, then the database work, partly because Stokes can uansfer the images to videodisc faster than the database can be set up. I was unsure about the real reasons for this, but they could be determine by talking to some of the people Stokes referred me to. As far as hardware goes, Stokes sounded like he doesn't really know very much about the Mac. He claimed that the Mac had no advantage over the PC in ease of use if the software was designed well, although I disagree with that rather strongly. Based on a number of years of working with novices on both systems, the PC is less intuitive and clumsier than the Mac when it comes to user interfaces. In any case, Stokes has developed a database package under Informix (which is Oracle compatible, or so he said) which runs on a number of different machines. If we decide to use Oracle, we could buy the Oracle package and then Stokes would provide us with his custom database for only the cost of support. This is curious because he could very easily create a stand-alone Oracle database and then just sell that (or give it away if he wanted) without the customer having to buy their own copy. Stokes wouldn't really comment on Image Premastering except to give me the name and 54 ASSIST Quarterly number of Bill Perry (202-857-7537) at National Geographic, which did independent tests of both Stokes and Image Premastering (and one other, actually whose name Steves did not mention). In addition, Stokes claimed that their quality has improved since then. Evidently, the American College of Radiology had 1 l"xl4" X-rays which Stokes claimed were optimized for the Image Premastering system and those came out better. As far as quality goes, any transfer from a positive image will lose quality in the transfer process, much as copying a tape or videotape loses quality. The contrast of a videodisc is usually around 20 to 25 with a maximum of 45, whereas a transparency is about 1(XX) and a slide about 250. Thus a great deal of contrast is lost when going to videodisc in any case. The transfer to a negative reduces this contrast lost by spreading out the contrast rather than clipping it, alUiough it also puts it through several generations of imaging. Stokes is aiming at a contrast factor of two to three times better than high definition video, which is as good as a monitor will get in the near future. He said that it is very difficult to match the original exacUy, and that matching the original better is their main task right now. Company: Turquoise Film/Video Productions Program: Videodisc services Hardware: NA Price: variable Turquoise said that they can provide anything up to a turn-key system. Their background is in motion picture processing, and they moved from that to providing software and hardware as well. Their database is an in-house one currently but it can import and export to a number of other formats. They are thinking about re-writing in Paradox, which can also run on a number of different machines. They charge an average of $2 per image, including hardware, and they can shoot either in St. Louis or on-site. They use a special motion picture film to get better quality images, but I have no sense how the quality of their images compares to the quality of either Stokes's or Image Premastering's images. It doesn't seem that there is anything remarkable about Turquoise in relation to the other two mastering services, but should the Archives decide to have a videodisc mastered externally to Cornell, it would be a good idea to talk more specifically to all three companies. Other Cornell Projects I spoke with Anne Camell about the project in the University Photography Department, and she said that they are having someone in Publications develop an in- house program. This p-ogram will catalog and store all of the information on their photographs, but it will also provide billing , usage, and reporting capabilities. They didn't think one of the commercial programs could provide all that, something which I doubt, given the power of some of these databases. They are looking at videodisc in the near future, for much the same reason as the Archives, perhaps in the next year or so. She didn't seem to have a wonderful grasp on what is entailed with the entire technology, since she didn't know about the problem with file sizes for scanned images, and she didn't know how many images could be stored on a videodisc. I also spoke with Dave Watkins, who is the head of Media Services and in doing so found my way back to the original Stokes project mentioned in the memos to and from Chris Pelkie. They are ciurenUy selecting slides, negatives, and prints for a free sample videodisc to be supplied by Stokes. Partly to test the quality of Stokes's service, they are trying to assemble a number of different types of images for inclusion. If the Archives wishes participate in this project to see how it works out, they should contact Dave Watkins. He is looking for up to three hundred of the most difficult type of images. The deadline for submission to this project is January 1st, 1990, which is fast approaching. Other departments that may be interested in this project include the Department of Entomology, which has 30,000 slides that they wish to use for diagnostic work, obviating the need to go to the slide collection itself. These slides must be of the highest quality because of their use and the fact that the disc will be sold to other universities. Plant Pathology and Veterinary Medicine may wish to do similar things. EvidenUy the Hotel School has a videodisc of wine labels and the Vet School and the Law School are still looking into the possibilities of some sort of image database on a videodisc. If the Archives wishes to look more closely at the specifics of a videodisc system in future, I strongly recommend that the department provide a number of photographs to Dave Watkins for this test project The project will provide a videodisc which can be shown to potential donors and with which we can test the pros and cons of various software packages. Such an opportunity should not be passed up lightly! I met with Margaret Webster, who runs the Architecture School's Slide Library. She has for some time been planning a videodisc project to keep track of the 350,000 slides in the library. Tliey hired an outside consultant, Nancy Humphries of ETECH, to research the possibilities and provide a system. Margaret has been putting records into the database and will be setting up the pilot project after the Slide Library moves in January Spring 1991 55 of 1990. They will be using a PC database called btricve/xtrieve (with which I'm not familiar because it was never compared in the literature with the more well- known databases) along with a specialized graphics board in the PC that will allow them to manipulate the images electronically. Manipulation of images is something which I did not explore particularly because of the expense involved, but is certainly a possibility fw the Archives. The question that must be answered to justify the cost of such a board is the use to which these photo- graphs are being put. If the photographs are ending up in publications designed and executed on a personal computer, or the images must be frequently manipulated, then a graphics board makes sense. However, if the publications in which these photographs appear use traditional methods of production, then additional 35mm negatives would be mwe useful. Margaret Webster knows more than most people on campus about vide- odisc systems because their system has been in the research phase for close to two years now. The final person from Cornell with whom I spoke was Mike Oltz from the Interactive Multimedia Group. He gave me some bits of information that may be useful. He thought that the Architecture School and the History of Art department were looking into something similar and were probably working along different lines. As it turns out. History of Art has dropped their project entirely, whereas the Architecture project is the closest to reality of any of the ones I've heard of. The exception to this is the Medical School, which has a system for pathology training using a Macintosh pseudo-database called Guide. They started out with no funding at all and now have close to five million dollars of computer equipment, which does lend hope to the Archives getting funding for a videodisc. The Interactive Multimedia Group has a sample videodisc from Image Premastering Services, which is currendy lost, but Mike will try to find it and let me know via email. He mentioned that Revlon, the makeup people, are also doing something like this. If we wanted to learn more we should talk to the advertising photogra- phy department There is no reason to assume that any of the other systems are better than a system the Archives could come up with, although the ability to perform information transfer might be useful in the future to avoid re-keying records. Similarly, standard information in each record would help in translating the records from another system when other departments wished to archive various photographs. Database information can be shared between PCs and Macs without too much trouble, so the specific machines used by different departments should not really matter, although the ability to transfer between the various databases matters. Cornell Coordination One problem that has come up time and time again in my research is that there are a number of Cornell departments working completely independently on similar videodisc projects. While such a lack of commu- nication is not unusual at Cornell, it is regrettable, particularly in a field such as this where the information really is fairly finite. It would be extremely useful if there could be a single person who would, if nothing else, have copies of all the various pieces of information collected by the different departments. That way, whenever anyone was thinking about starting such a project, the information would be more or less at hand and would include names of the people at Cornell who are good resources. This person would merely dissemi- nate information and would refrain from making any recommendations as far as hardware or software go in order to avoid the politics. The Interactive Multimedia Group would seem to be a logical group to coordinate or various videodisc information, but because they exist completely on soft money for specific projects, they are not set up to handle any sort of coordination. Geri Gay said that Media Services was one place coordination could come from, and some part of CIT Services would be another. People to talk to in CIT include Larry Fresinski, Donna Tatro, and if all else fails, Stuart Lynn. Another area in which the various departments could pool resources would be in setting up facilities at Cornell for transferring images to videodisc. I gather that there are close to a million images at Cornell that could be put on videodisc if the process was cheaper and easier. Dave Watkins in Media Services is the person to talk to about such a project Margaret Webster in the Architec- ture Slide Library would also be very interested. Other Non-Cornell Projects I've found names of people at other institutions who have done something along these lines or are thinking about it Talking to them might help the final decision because you can get an opinion from someone in a similar position. I did not get more detailed informa- tion from these people since it is often easier in this situation to use academic channels for sharing informa- tion, and the Archives ah-eady has contacts in some of these institutions, whereas I would be going in cold. So it doesn't make sense for me to talk to everyone immedi- ately unless it seems that they have something important to offer to the decision-making process right now. That step can come if and when the Archives decides on a specific system or type of system. If I know of a way of contacting the jjeople below, I've mentioned it Most of this information comes via electronic mail, so I can ask for additional contact information if desired. My apolo- gies for the lack of organization, but no method proved 56 ASSIST Quarleriy itself better than formatting for maximum readability. Elizabeth Wood mentions joint project between the Emergency Medicine and Radiology l5epartment of Los Angeles County and the University of Southern California Medical Center that had their mastering done by Image Premastering. Elizabeth H. Wood Computer Services Librarian Norris Medical Library University of Southern California ewood%phad.hsc.usc.edu@ usc.edu The AV Department of Hombake Library at the University of Maryland is developing a videodisc in conjunction with the National Agricultural Library. 1 wonder if this is the Fwestry Service collection men- tioned above. David Austin mentions two other projects. First, Andrew Eskind at the Eastman House is working on something to do with a videodisc. Second, Jim Sheldon at the MIT Media Lab is working on a videodisc of Edweard (sic) Muybridge motion pictures in conjunction with the Addison Gallery of American Art, Phillips Academy, Andover, MA, 01810. Finally, David says "Also, make sure you check the SN/G: Report on data processing projects in art (1988). It is not yet on-line but available in hard copy, maybe even at Cornell. It is a list of projects registered with the Scuola Normale Superiore, Pisa, Italy and the Getty Art History Information Pro- gram, Los Angeles." David Austin U29716@UICVM Jim Sheldon jls@ media-lab.media.mit.edu The AVIADOR (Avery Videodisc Index of Architectural Drawings on RLIN) project at the Avery Architectural and Fine Arts Library at Columbia sounds very similar lo what the Archives might want to do. In addition, RLG is working on a way of hnking a videodisc to an RLIN terminal, which would be very interesting. Janet Parks sent me a copy of their literature on the videodisc system. Janet Parks Curator of Drawings Avery Architectural and Fine Arts Library Columbia University New York, NY 10027 212-854-6738 Jane Kleiner mentions several videodisc projects. one of which is the Emperor I collection done by Ching Chi Chen at Simmons. It is quite sophisticated and includes sound as well. She thinks MIT has an architec- tural collection on videodisc and adds that the National Agricultural Library has a collection of historical photo- graphs from the Forestry Service on videodisc. Jane Kleiner notjpk@lsuvm Lennie Stovel mentions that there would be more information in the Library of Congress* literature on their Prints and Photographs Division's videodiscs, although he does not give a specific contact. Lennie Stovel Library Systems Analyst Research Libraries Group bl.mds