🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: vmsnet.internals
Article #275

re: PATCH on ALPHA ?

#275
From: George J. Carrette
Date: Thu, 03 Sep 1992 13:09
65 lines
3336 bytes
In article <715417768.652976.MILLER@TGV.COM>, MILLER@TGV.COM writes:
> Has any one else here had any problems using Patch?

Yes. I love the idea of patching things, especially patching running
systems. The good experiences comes from my LISP/LISPMACHINE days.
And I beleive that -as-a-model- patching is the correct way to do 
software support and upgrades.

So, when we are patching parts of our system written in 
imbeded/interperted language we send out stuff that is indeed a patch.
However, for stuff compiled with the C compiler, I found that I could
never seem to be able to take the time to convert a new procedure definition
changed at the source level into a convenient set of PATCH instructions.

Why, because our product is split up into about 39 different images,
with a typical size of 500 kilo-bytes. And a 500 kbyte VMS image compresses
with uucp_bin:compress into about 384 kbytes. And with a V.32 modem we
can UUCP that file in about 10 minutes.

Then, well, things like UID files (output of the MOTIF UIL compiler)
also end up making a significant portion of the size of the "binary"
stuff in a product. And that isn't even machine code.

--- WHAT THE WORLD NEEDS NOW: ---

A good "arbitrary-binary-file" diff/patch utility. Something with
good compression and error detection. Perhaps using technology
similar to what high-definition television would use to compress
differences between frame image[t] => image[t+delta(t)]

Something that could handle the addition of a few bytes in the
middle of an image, and other changes, without generating
grossly large differences.

> On those occations when I need to read someone a patch over the phone
> I have them read it back to me before they perform an update. 

Unfortunately, the only thing we can patch like that (via a human) is very
small chunks of interpreted code. But it is always risky.

> With Patch we can have a fix out to the customer within an hour
> instead of the day or two it takes to ship a new tape.

Indeed, fixes-while-U-wait is the only way to do good customer support.

We end up putting a good deal of work into getting the EMAIL connections
going. Boy is that a hassle. Usually the customer is already connected
to some internal DECNET that covers the whole company. But they don't
know that they have internet mail capability, usually through some
obscure gateway run by the research division or some non-responsive MIS
group in central administration.

For other sites, e.g. a Government lab, if we are lucky they can find
an IBM-PC someplace with a modem and kermit.

It can be a hassle, but we find that is some of the most hassle-intensive
organizations you have problems like it taking 2 days for a FEDEX-delivered
package to make it from shipping/receiving or the mail-room back into
the depths of an organization.

Anyway, to grease the administrative wheels, we actually CHARGE EXTRA
for a SUPPORT CONTRACT that doesn't include an EMAIL or DIAL-UP capability!


-gjc
Technical Details
Message-ID: <2600@mitech.com>
Path: arc1-ca.pugleaf.net!sparky!uunet!mitech!gjc
References: <715417768.652976.MILLER@TGV.COM>

Show Full Headers
Complete Article Headers
Path: arc1-ca.pugleaf.net!sparky!uunet!mitech!gjc
From: gjc@mitech.com (George J. Carrette)
Newsgroups: vmsnet.internals
Subject: re: PATCH on ALPHA ?
Message-ID: <2600@mitech.com>
Date: 3 Sep 92 13:09:40 GMT
References: <715417768.652976.MILLER@TGV.COM>
Organization: Mitech Corporation, Concord MA
Lines: 65