Date: Tue, 28 Jan 92 18:43:06 PST From: rab@sprite.Berkeley.EDU (Robert A. Bruce) Subject: Re: cdrom publishing info, please Here is a description of some of the things I learned while producing two cdroms. The first disc contains the X11R5/GNU source code. The second disc contains the Simtel20 MSDOS Archive. If you are thinking about making a cd, I hope you find this information useful. Equipment: The most important thing you will need is a big magnetic disk. You could do it with a 700 meg disk, but it is better to have twice that amount so you can have two copies. That way you can make changes on one copy, and be able to recover from the other if you screw something up. You will need a tape drive for backups, and for shipping your files to the disc publisher. I used an 8mm Exabyte. A 4mm DAT will also work. All the disc publishers I talked to will accept either 4mm or 8mm. Make sure you always make a backup before you make major changes to your files. It is possible to set up all your files under MS-DOS or MacOS, but I wouldn't recommend it. Unix provides much better tools for working with big filesystems. I used a sparcStation-1 with a 1.6GB Wren-8 disk drive. I had absolutely no problems with either hardware or the system software. It is good to have a couple different types of computers and CD drives to test things on. I tested my stuff on a sparcStation-1 running SunOS with a Sun CD and a 386 running MS-DOS with a Magnavox CD. These systems have different byte orders, different OS's etc. I figured if things worked on both of them, it would probably work on just about anything. Handling data: You need to get used to thinking in big numbers. My X11R5/GNU CD-ROM contains over 35,000 files, and about 650 Megabytes of data. It is almost impossible to do anything by hand. You need to be able to automate things. The Unix `find' command is very useful for traversing directories and performing operations on files. Choosing the format: The first big decision you need to make is what format to use. If your CD is targeted for just one platform, then you might be able to use a special format. But if you want your CD to be useful on different types of computers then you are stuck with ISO-9660. There are a couple levels to the ISO-9660 standard. Most of the comments here apply to level one, the most generic level. Besides portability, another big reason to use ISO-9660 is that it is the only format supported by many of the publishers that press discs. The ISO-9660 format places the following restrictions on your filesystem: The depth of directories is limited to eight levels. This wasn't much of a problem for me. Filenames are limited to eight single-case characters, a dot, and a three character extension. Filenames cannot contain special characters, (no hyphens, tildes, equals, or pluses). Only single case letters, numbers, and underscores. Directory names cannot have the three digit extension, just eight single-case characters. I didn't know about this restriction until after I sent first my tape to the publisher. They called me up and said the mastering software was choking on my directory names. I had to make a new tape and send it off to them. This problem delayed the first disc by about a week and a half. Here are some examples of legal and illegal filenames: Legal Illegal Why test1c.txt test-1c.txt hyphen test_1c.txt test 1c.txt space test.1c test.1c.txt more than 1 period readme Readme not single case If you are naming your own files, these restrictions are not too onerous. But if you are trying to shoehorn a system like X11R5, or Gnu where the files already have unix-type names, then it is more of a problem. I wrote a program to scrunch all the filenames, and then created a file in each directory that maps the new name onto the original name. I then wrote a program to recursivley copy a directory, (or, optionally, create a directory of symbolic links) using the original names. If any of you ever need a similar program, you are free to use mine without any restrictions. The source code is included on the X11R5/GNU disc. The Simtel20 disc was much easier to convert to the ISO-9660 standard. Since the filenames already conformed to the 8+3 MSDOS format, all I had to do was fix a few files that had hyphens or dollar signs in their names. Level two ISO-9660 allows longer filenames, up to 32 characters. But many of the other restrictions still apply. Level two discs are not useable on some computers, especially PC's running MS-DOS. Disc manufacuturing: I had my X11R5/GNU disc made by Discovery Systems 1-614-761-2000. I shopped around a little before deciding to go with Discovery. They didn't have the absolute lowest prices, but they seemed to be more computer oriented. Most other CD companies do mostly audio CD's for the music industry, and cut a few CD-ROMS on the side. But Discovery had some people that specialized in just CD-ROMS. They were able to answer my technical questions with out any problems. My second disc, the Simtel20 disc, was published by Disc Manufacturing Inc. 1-800-433-DISC. They also have people dedicated to CD-ROM production. Overall, I think DMI did a better job than Discovery. Their prices are better, they stick to a tighter schedule, and in my opinion their quality control is better. For my third disk, I am planning to go with DMI again. Another company that has some good recommendations is Optical Media International, (408) 376-3511, omi@applelink.apple.com. If anybody has had CD's made by OMI, let me know how it went. The publisher can accept your data in several different forms. You can send them a `one-off' (see below), or a tape. They will accept 8mm videotape, 4mm DAT, or 6250bpi 9-track tape. I used 8mm. The tape can be in several formats: tar, ANSI format, or binary image. I used tar because that was the easiest for me. In order to send them a binary image you have to have your own pre-mastering software. I looked into doing this. The Rockridge System by Young Minds 1-714-335-1350 seemed like the best, but at $6995 it was too expensive for me. Rockridge provides extensions to the ISO-9660 format that allows full Unix filenames to be used. The Rockridge extensions will probably become more widespread as more drivers start supporting the extensions. When you send your tape to the publisher, it is a good idea to have the following statistics available. It will enable them to layout your files more efficiently: Maximum number of files in any one directory Total number of files Total number of directories Average file size Largest file size I sent two duplicate tapes. If there are any errors on the first one, they can switch to the second. All of my tapes worked the first time, but if there had been problems, the backup tape would have saved lots of time. One-off: A one-off is a single copy of a CD. You can get one made from a tar tape for about $300. You can then test it and make sure everything is correct. You can skip this step if you want to, and just ship the tape directly to the publisher. I had a one-off of my X11R5/GNU disc made and I was glad I did because several things were screwed up. I corrected all the problems, and then I sent the tape to the publisher without having a second one-off made. I had my one-off made by `On-Site CD' (408) 867-0514. They are a small company, just two guys working out of a spare bedroom. But they are very quick. I drove down to Saratoga and dropped off my tapes. They FedExed me the one-off two days later. There were some problems with the one-off, mostly my fault, but some of them were their fault. But they had just started doing business a few weeks earlier and I was one of their first customers. I am sure they have ironed out the problems. JVC is coming out with a relatively cheap ($12500) one-off system that should reduce the cost of having a one-off made. I will have one of these units for testing by March or April, so if you want a one-off, or very low volume cd-rom production, let me know and I might be able to help you out. Misc. problems: There is a lot of buggy software in this world. Even if your own software is reliable, the software used by the people that make your discs may not be. Most software handles the common cases properly but often does not handle unusual cases well. All of these things caused me problems: Filename that start with a dot (e.g. `.foo'). They get left behind if someone does a `mv *.*'. Zero length files. Some mastering programs will not create them. Read only files on a tar tape. At least one tar program out there will `creat' the file using read-only mode, and then try to open it for writing. The write fails, and you end up with an empty file. Make sure your files are mode 666 (or 777 if they are executables). The mastering software will make all your files mode 555 when it puts them onto the CD, regardless of what their original mode was. Empty directories. Some mastering programs do not create the directory until they put the first file into it. So empty directories never get created. All of these problems are pretty easily eliminated. Files that start with a dot, zero length files and empty directories are usually just cruft that should be eliminated anyway. Disc Label: The publisher will send you the precise dimensions for the disc label. You will need to create a film positive, emulsion side up. You can use 2 colors in addition to the silver background. If you use more than two colors you will have to pay extra. Make sure your artwork doesn't bleed off the disc. I don't know anything about this art stuff, so I hired a local graphics artist to do it all for me. I just gave her a rough sketch of what it should look like. She charged me $125, and did an excellent job. She even drew a pretty good picture of a gnu. Packaging: The publishers provide several different packaging options. The cheapest is to get no packaging. You just get a box of discs. For an extra 35 cents you can get each disc placed in a jewelbox, which is the same plastic case that audio cd's are kept in. You can then have the jewelbox either shrinkwrapped, put in a blister pack, or put in a long cardboard box like most audio cd's are sold in. I had my discs put in jewelboxes, but did not have any further packaging done. If you want to put inserts into the jewelbox, you need to provide them to the publisher. They will send you the specifications and dimensions, and then you need to work with a local printer to make them. You send them with your order, and the publisher will insert them into the jewelbox for a small cost. I did not use an insert with my discs. Cost: Here is how I figure my expenses for my first disc: One-off: $20.00 Gasoline (to deliver the tape) $300.00 One off production $30.00 Overnight Fedex --------- $350.00 total for one-off Mastering: $500.00 partial mastering $700.00 mastering $875.00 pressing (500 x $1.75/disc) $174.00 shipping for 500 discs 2nd day UPS ------- $2149.00 total for mastering Misc: $125.00 Art work for disc label $120.00 8mm tapes (backups, shipping files) -------- $245.00 $350.00 one-off $2149.00 mastering $245.00 misc -------- $2744.00 total This total does not count the cost of equipment because I can use it again, and it does not include the cost of my time (probably about 100 hours or so). The mastering had to be done twice because I screwed up the directory names. The $700 was a special ``first-timer'' introductory price. These first-timer discounts are pretty standard, and many publishers offer them. For regular customers, the cost is about $1400 for a standard five day turnaround. Cheaper rates are available if you are willing to accept longer turn around times. The discs were shipped to me in five boxes of 100 discs each. I was in a hurry to get them, so I had them shipped 2nd day air. In hindsight, I could have saved some money by shipping only one box by second day air, and the other four by regular ground delivery. That way I could have had the first 100 to test out, and meet the first orders, and then received the other 400 a week or two later. The total cost was only about half as much as I had expected to spend. I had about $5000 that I saved up for a down payment on a new pickup truck. I figured I could get by with my old clunker for a little while longer, but my girlfriend was really pissed off when I told her I was spending the money to make a cd-rom. I was happy that it turned out to cost less. Here is how I figure my expenses for my second disc: $16.00 8mm tapes $125.00 Artwork for disc label $18.00 UPS next day $750.00 Mastering + 50 discs + shipping ------- $909.00 total for first 50 discs $125.00 Remount fee $750.00 pressing (500 x $1.50/disc) $40.00 (est) UPS Ground -------- $915.00 for additional 500 discs Total: $1826.00 for 550 discs Time: The first disc took me about two months from start to finish. Here is a rough schedule of how long each phase took: week 1 -- collect information week 2 -- organize filesystem, munge filenames, create index files, lots of testing, compiling, etc. week 5 -- One-off made week 6 -- fix problems, more testing week 7 -- sent the first tape to the publisher week 8 -- fixed directory names, sent the second tape to the publisher week 9 -- discs are done My second disc went a lot smoother. Partly this was because I knew what I was doing, but mainly it was because I pretty much just slapped the archive onto the disc `as is'. There was very little processing that had to be done. I started right after Christmas break, and I was done before the end of January. week 1 -- Download archive week 2 -- Edit index files, lots of testing week 3 -- Sent the tape to the publisher week 4 -- discs are done ---------------------------------------------------------------------------- Making a CD is not difficult. If you have been thinking about it, I recommend that you go for it. I am currently producing several more. If you have any questions about making a CD, feel free to contact me. If you need any help in putting your CD together, I will be glad to help. I have several large discs and tape drives, and I hope to have a one-off unit soon. ---------------------------------------------------------------------------- If you are interested in CD's, I still have X11R5/GNU CD's available. The price for one disc is $40. Additional discs are $20 each. All the GNU source code All the X11R5 sources, including all contributed software All the comp.sources.x archives Sparc binaries and libraries for the Gnu programs, and the X11R5 server and clients. I also have cdroms containing the Simtel20 MSDOS Archive. It has about 420 Megabytes of utilities, programminmg tools, source code, technical documentation, etc. If you want to see exactly what is on the disc, just look at the MSDOS directory at wsmr-simtel20.army.mil or look in the mirrors/msdos directory at the mirror sites wuarchive.wustl.edu or oak.oakland.edu. The price for one disk is $25. Additional disks are $12.50 each. Shipping and handling is $5 U.S/Canada, and $10 for overseas. S&H cost is per order, not per disc. If you live in California, you need to add 8.25% sales tax. Here are some other discs I am planning to make: 1. Misc. source code disc. 2. Graphics software disc. 3. Technical documentation disc. 4. Desktop publishing software. 5. OS software disc. The source code disc is expected to be done around 1 March 92, and will cost $40/$20. There is no release date yet for the other discs. ---------------------------------------------------------------------------- Bob Bruce 1547 Palos Verdes, Suite 260 Walnut Creek, CA 94596 1-510-947-5996 rab@sprite.Berkeley.EDU