gnu@hoptoad.UUCP.UUCP (08/12/86)
Two recent proposals for new file transfer protocols remind me of the
ongoing lack of vision among the micro community.
Multitasking, interrupt driven operation is a reality in many machines,
and will be standard in the future, but people are still building
single threaded, half duplex, master/slave protocols. It should be
possible for multiple tasks to "talk" back and forth through a
protocol, rather than just dumping a file on the remote system; imagine
reading a BBS over the line, and while you are reading, a file is
being transferred in the background over the same phone line.
Automatic forwarding of files and messages is clearly useful in the
Usenet, Fidonet, and many other existing networks. But the micro protocol
proposals all tend to assume that there is a human at one (or both!)
ends supervising the transfer.
Error-free operation is useful when humans are involved and critical when
they are not, but the proposed protocols continue to rely on unprotected
single control characters as acknowledgements, simple failure-prone
setup handshaking, and ignorance of flow control issues ("somebody else
will handle that").
While certainly minicomputer and mainframe protocols are nothing to shout
about, they at least haven't ignored ALL of the above lessons.
I recommend that the micro world stick with what they've got rather than
create yet another set of incompatible protocols that will need to be junked
just as they are gaining acceptance. Rather than implementing half-baked
"improvements", we could spend the time to design a real protocol that
will be useful for more than a few years.
I worked in the PCNet Project of Peoples' Computer Company circa 1977.
You might have read about it in Dr. Dobbs Journal. We designed and
implemented a layered protocol that would work in half-duplex, in
BASIC, and tranfer anything over a link that could only pass uppercase
letters and digits -- but would extend compatibly to full-duplex binary
operation by just changing the bottom level and keeping all the
application software the same. I can post or mail specs of the
protocol if anyone is interested. The [volunteer] project fell apart
before our stuff was shippable, and Kermit took up the slack. Even
though it's an old design, it's better than the "windowed xmodem" and
"fast" proposals; had it been released in 1977 it would still be current
today.WANCHO@SIMTEL20.ARPA (Frank J. Wancho) (08/12/86)
The archives of the PCnet project are available via ANONYMOUS FTP here on SIMTEL20 in PD:<PCNET>. --Frank -------