[ The following message is a message is a side conversation
[ between Zoltan Janosy and Robbie Rhodes which is being shared 
[ with us.  The message, forwarded to us by Robbie, is a message
[ from Zoli with Robbies annotations.  Normally I reserve the "["
[ to mark annotations I've made, but it seems to work in this case.
[ All lines in this message after this comment block with "["'s are Robbie's.  
[ Jody

=====atch======

From: "Zoltan Janosy" <janosy@hit.bme.hu> 
To: rhodes@foxtail.com (Robbie Rhodes)
Date:          Thu, 30 Nov 1995 15:03:03 GMT+1 
Subject:       Re: Roll file formats

Robbie, [regarding:]

> Jody and Zoli, I've asked Wayne Stahnke for his input.  Meantime,
> would you please each write a letter (responding to Walter Tenten's 
> note on this subject) about YOUR ideas for the file formats
> appropriate to storing the raw, unedited data that comes 
> spewing out of an optical scanner.

We use standard TIFF format with the CCITT Group 3 compression. 
This is suitable only for black and white images, though. After 
processing we create a .prl file, which is in fact a MIDI Format 0 
file having an additional information field (a sequencer specific 
meta-event). We had problems with this format. Some sequencers 
(namely Cakewalk) couldn't open the file, or got some garbage.

[Cakewalk Pro is reportedly much improved.  -- Robbie ]

We use the Standard Midi Files (SMF) meta-events for copyright, title, 
labels, text, etc. I think it is always good to stick to an already 
established file format than to create a new one. E.g., Wayne's file 
format is fine, but I don't see the advantage of it. He said it is
the note limitation of MIDI (127 notes per channel) and the maximum 
possible event delta-time that limits the SMF format. Well, one can 
use 16 channels per track and with Format 1 files even several tracks. 
For example one would use a separate track for each of the instruments 
of an orchestrion. The variable length encoding used in
SMFs [Midi files] gives IMHO [the time axis] more than enough 
resolution. And one can make a very long pause with extremely 
high resolution by adding some dummy events in between. These are 
not really serious limitations.

[ Zoli is referring to Stahnke's file format intended primarily
[ for music roll master files.  Wayne realized there is a need for 
[ more than 128 channels in large organ formats, and so he uses
[ a simple format that supports 240 channels plus 15 meta-channels 
[ which are used for annotation (Title, Performer, etc).
[ Otherwise, Wayne's "perforator master" format resembles 
[ closely Richard Tonnesen's perf control format.
[
[ I disagree with Zoli on the time-axis resolution.  Midi files 
[ are strongly oriented toward "ticks per measure", whereas
[ the scanner output is only oriented by distance.  Distance
[ and Midi-timing are not equated until after much post-processing. 
[ TIFF format allows the use of any arbitrary resolution on
[ both axes, expressed in decimal inches or centimeters. 
[ -- Robbie

For storing the image itself I'd suggest a high-speed magneto-optical 
disk. One problem with such media is that their guaranteed lifetime 
is somewhat limited (about 50 years - funny that paper rolls are 
already a hundred years old). Maybe writable CD would be better, but 
I'm not sure. I've heard about oxidation of the substrate surface 
making CD-s unreadable in a couple of tens of years !???

[ Think about the durability of old glass-plate photographs! ]

For compressing I'd vote for a lossless compression scheme (no JPEG). 
There are thousands of different graphical file formats. As I said we 
use TIFF because it is available on all platforms, but maybe something 
else would be better (portable pixmap, maybe). One thing that should 
be kept in mind is that it should be portable across different operating 
systems. I wouldn't choose something that is Windows or Mac specific. 
Greyscale would be fine in my opinion for all purposes, but maybe 
B&W would suffice in many cases (if there is nothing written on the 
surface of the roll).

Well, that is what I think. I vote for standard, established file 
formats. That will make the files readable in the future as well 
(if there will be devices to read them!). Yes, this is really a 
problem, since the computer industry changes so rapidly, that there 
is no guarantee that electronic media will be readable in hundred 
years from now. CD seems to be a solid format, at least given the 
amount of existing CD-s. But remember, in ten years it will be 
almost hopeless to find a vinyl LP player (can you replay 
phonograms easily?). So it is possible that the current CD 
format will change. Hmm, tough.

Please forward this letter to the others involved, and I'd 
like to get the others responses too. Maybe we should talk on 
the mailing list.  [Yes.]

Best regards,

Zoli

= (Mr.) Zoltan Janosy <janosy@hit.bme.hu> ¶
= Technical University of Budapest¶
= Department of Telecommunications¶
= Sztoczek u. 2. 308., H-1111 Budapest, Hungary ¶
= Tel: +36-1-463-2093, Fax: +36-1-463-3266

[ I think TIFF format with CCITT Group 3 compression is related to 
[ the current standard for facsimile transmission.  It sounds like 
[ it's quite appropriate for the raster-scanned image, and it is a
[ lossless compression scheme.  Has anyone tried viewing a long file 
[ like this with a graphics viewer/editor?
[
[ We still need suitable formats for the intermediate processing steps. 
[ Wayne, what do you suggest?

[ -- Robbie