Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!uunet!hoptoad!gnu
From: gnu@hoptoad.uucp (John Gilmore)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Fax over tcp/ip
Message-ID: <8734@hoptoad.uucp>
Date: 19 Oct 89 06:54:15 GMT
References: <8910131208.AA18264@iapetus.rice.edu> <2078@brazos.Rice.edu> <30@van-bc.UUCP>
Organization: Grasshopper Group in San Francisco
Lines: 40

sl@van-bc.UUCP (Stuart Lynne) wrote:
> I would suspect that if a fax transfer standard is implemented it should 
> involve a handshake at the beginning where each end tells the other what 
> it understands. Then if you have a format which the other end doesn't 
> understand you know you have to convert before sending otherwise you just 
> send it.

Clearly what we want on the Internet is more like what computer fax
over phone lines should be doing (except that the designers of fax
didn't think about extensibility, and the people who implement it are
too busy (hacking slimy binary MSDOS software) to support real
computers or consider the larger issues).  You should hand it a
document in ANY format; it sends it to the recipient in the recipient's
choice of format.  E.g. you hand it troff source; if the recipient can
handle troff source, it gets sent that way; backoffs to ditroff
intermediate form, PostScript, MacDraw, HPGL, bitmaps, G3 and G4 fax,
etc, are all possible depending on what the recipient can handle.  This
has to be negotiated separately for each recipient, since one could be
a phone-fax relay service and another a window system or something.

Note that data modems have the same sort of problem:  when answering a
call, the two modems need to negotiate to decide what mutually agreeable
protocol will be used over the line.  Current modems use a horrid solution:
try one, wait a while, try another, ...  burning billable time at your
expense.  A better solution would have been to send a burst of touchtone
digits, each digit pair or triple indicating a supported protocol.
The other side would send its best choice among them and it all happens in a
second or two.  (This protocol is stolen from the uucp startup sequence.)

For tcp/fax negotiations I'd define a standard set of format names in
ASCII, rather than using numeric codes; it's easier to use and extend a
set of names.  Also, we probably want to send more information during
the negotiation, like the file name and size, sender's identification,
etc, so the recipient can use the info in deciding what format to
request (e.g. if I know that I get a message weekly from you, and it
works best in PostScript, I can set up to request PostScript from you
even though you are supplying it in TeX).
-- 
John Gilmore      {sun,pacbell,uunet,pyramid}!hoptoad!gnu      gnu@toad.com
		"Watch me change my world..." -- Liquid Theatre