Computer historian? [LWN.net]

Computer historian?

Posted Mar 23, 2007 11:46 UTC (Fri) by man_ls (subscriber, #15091)
In reply to: GNU/Busybox ?!? by landley
Parent article: The road to freedom in the embedded world

Who _I_ am (I'm not nigelk but he was responding to my message) is among other things, a computer historian (see http://landley.net/history/mirror for some random snapshots of research materials).

You consider yourself a historian? Then maybe you should set your academic standards a bit higher. Random examples from your earlier messages:

Back in the 1970's there _was_ no proprietary software market to speak of.

In 1978 Microsoft had 11 employees, all doing proprietary software. It was the year when WordStar was published, and VisiCalc was soon to follow. But this is just in the microcomputer area; a fledging 1970s market saw the birth of such companies as Compuware (1973), Computer Associates (1976), SAS Institute (also 1976) or Oracle Corporation (as SDL, 1977); meanwhile Software AG had been founded in 1969 in Germany; and in 1972 in the US. I would hardly call that "no proprietary software market".

Micro-soft was one of the first, and they were a shoestring operation with three employees, one of which was part time and one of which had a day job at HP.

As someone has pointed out, Microsoft had no HP employees. Paul Allen used to work for Honeywell, Bill Gates dropped out of school. You are confusing Microsoft with Apple.

The _very_ first might have been Digital Research, which was a one-man operation Gary Kildall ran out of his living room to make and sell CP/M for the very first commodity hardware platform, the S/100 Altair clones.

DRI was hardly the first; by 1974 (the year DRI was founded) several companies were successfully selling business software. By 1980 Apple had 1000 employees, while Digital Research had more than 200. It never was a "one man operation"; from the start (1974) his wife was an integral part of the company.

AT&T didn't try to commercialize Unix until 1983, by which time it was about as old as Linux is now.

According to Levenez, by 1983 there were already several commercial Unix variants, including Microsoft's Xenix and HP-UX. Presumably AT&T were making money from it, which counts as "commercializing" IMHO.

Stallman's been marginalized because he hasn't done anything new for 15 years.

Stallman has been marginalized? Well, I guess that going around the world speaking about Free Software in various public (e.g. the European Commission) and private venues, and publishing books and articles all around doesn't count. Meanwhile his GNU project has started subprojects such as the GIMP or GNOME, and his GPLv3 committees have brought together most companies involved in Free Software from around the world. Not bad for a marginal character.

And the first complete reimplementation of Unix (BSD, again predating the GNU manifesto) still has several forks active today.

As has been mentioned before, not true: Coherent and a few other independent variants existed before BSD itself was independent.

The big advance in open development in 1984 was the invention of the program "patch", which was done by Larry Wall (who went on to invent Perl). What Stallman had was an FTP site donated by MIT, back when that was hard to get, so lots of people like Wall signed up to get distribution on ftp://athena.ai.mit.edu. Stallman claimed credit for this code but he had nothing to do with it, he was running the sourceforge of his day.

I have just downloaded a fresh copy of patch and there is a very clear explanation of the roles of Larry Wall and the FSF. Without an explicit reference it is impossible to check your statement.

When FTP space became easier to get (cdrom.com and sunsite were both pretty active by the early 90's) the GNU project faded into well-deserved obscurity because they couldn't browbeat people into putting up with Stallman as a condition of getting distribution for their code anymore.

I'm not sure "well-deserved obscurity" describes particularly well their past or current status. By the time I got acquainted with Solaris in 2000, the first thing everyone did to accomplish anything useful was to download several GNU packages such as Bash or GNU tar; not to speak about GCC. I am certain that GNU's popularity was not because it provided "FTP space".

The important projects the FSF once maintained all stagnated and forked, gcc->egcs (and the name was handed over with gcc 2.95), glibc->glibc2 (and Ulrich Drepper who forked it and still maintains the fork was kinda pissed when the FSF tried to muscle back in on it: http://sources.redhat.com/ml/libc-announce/2001/msg00000.... )

GCC has not stagnated; it is currently past its 4.0 release and moving along nicely, having been redesigned several times since the fork you mention. Similarly GlibC: the differences you mention have been put aside and development seems to move along. Many other GNU projects (such as the aforementioned GNOME or the GIMP) are moving along at constant speed. Your statement is so biased that it is hard to find any GNU packages which have actually languished or stagnated because of political differences, any more than e.g. your ex-project Busybox.

I could go on. Unverifiable and biased statements are not (or should not be) the modus operandi for a historian. You write well, but your foundations are really too shaky. Please verify your statements before posting them in public; even a few trips to the Wikipedia can save a lot of embarrassment later.

(Log in to post comments)

Computer historian?

Posted Mar 23, 2007 17:47 UTC (Fri) by landley (guest, #6789) [Link]

> You consider yourself a historian?

It's a hobby, and your argument here seems to be that there were perhaps
a half-dozen low-volume examples about 18 months before the turn of the
decade. I knew about this. That was the "to speak of" part.

> > Back in the 1970's there _was_ no proprietary software market to
> > speak of.
>
> In 1978 Microsoft had 11 employees, all doing proprietary software. It

Yup. Which they sold to hardware manufacturers for bundling with their
product, because selling directly to end users was not a commercially
viable option. Right at the end of the 1970's the bespoke software
development market finally started to scale.

Microsoft big customers were computer manufacturers like Mits and IMSAI,
their sales to end-users sucked badly enough to prompt the famous "letter
to hobbyists" in 1976. Microsoft's big cash cow as late as 1980 was its
contract for TRS-80 ROM images, which Gates talks about at some length in
this 1980 audio interview:

http://landley.net/history/mirror/ms/gates.mp3

According to the book "On the Edge"
(http://www.amazon.com/Edge-Spectacular-Rise-Fall-Commodor...)
Microsoft sold commodore unlimited rights to ROM BASIC for a one time
payment of $10K, and went on to sell it to Radio Shack for $20k. (This
was something like 1979, I'd double check and cite a page number but the
book's at home and I'm not.)

I already mentioned that Paul Allen's day job was at Mits, not HP. This
is off the top of my head, I'm mostly not checking references for this
thread.

> DRI was hardly the first; by 1974 (the year DRI was founded) several
> companies were successfully selling business software.

For mainframes and minicomputers, sure. I mentioned the bespoke bit
where you comission very low volume software at extremely high per-unit
costs?

> By 1980 Apple had 1000 employees

Selling hardware.

> was the year when WordStar was published,

http://www.wordplace.com/ap/ does a decent job of covering that (starting
in chapter 3), but again: volume in the toilet back on CP/M.

> and VisiCalc was soon to follow.

Which I mentioned.

> But this is just in the microcomputer area; a fledging 1970s market saw
> the birth of such companies as Compuware (1973), Computer Associates
> (1976), SAS Institute (also 1976) or Oracle Corporation (as SDL, 1977);
> meanwhile Software AG had been founded in 1969 in Germany; and in 1972
> in the US. I would hardly call that "no proprietary software market".

This was the bespoke market I mentioned. Each copy of that software sold
for a year's salary of programmer time, because the total number of
machines it could run on was so limited. With a setup like that, each
copy is essentially tailored to that customer.

> According to Levenez, by 1983 there were already several commercial
> Unix variants, including Microsoft's Xenix and HP-UX. Presumably AT&T
> were making money from it, which counts as "commercializing" IMHO.

I've read his chart and exchanged email to the guy. Xenix was
commissioned in 1979 (SCO did the implementation as a two-person
consulting shop, and that was founded in 1979). And this gets us back
to "software bundled with a hardware purchase" again.

And AT&T were _forbidden_ from making money from it due to their 1959
antitrust consent decree. They were a regulated monopoly and could not
diversify out of the telephone business. They agreed to be broken up in
1983 (the breakup was in 1984 but the judgement was in 83) to get out
from under that antitrust decress so they could diversify into things
like the computer industry. (Bell labs came out with the transistor, the
laser, and unix, and all they could do with any of it was upgrade their
phone switches.)

> > And the first complete reimplementation of Unix (BSD, again predating
> > the GNU manifesto) still has several forks active today.
>
> As has been mentioned before, not true: Coherent and a few other
> independent variants existed before BSD itself was independent.

BSD was clearly started before coherent, but coherent was finished first,
therefore it's "first". GNU was started before Linux. GNU still hasn't
been finished. Therefore GNU is first.

Pick one, will you?

> By the time I got acquainted with Solaris in 2000, the first thing
> everyone did to accomplish anything useful was to download several GNU
> packages such as Bash or GNU tar; not to speak about GCC. I am certain
> that GNU's popularity was not because it provided "FTP space".

In that case it's because Ed Zander decided to unbundle the compiler (and
other things) from Solaris so he could charge extra for it, and this made
gcc the de-facto compiler of Solaris. I mentioned this in another post.

(Personally I lump Solaris in with Desqview and OS/2 as "of only
historical interest", but I realize there remains a vocal minority who
will defend it for years to come. Just as OS/2 had.)

> GCC has not stagnated;

The original development line did. Cygnus forked egcs and took over the
name, and these days the driving force behind it seems to be
CodeSourcery.

I mentioned this.

> Unverifiable and biased statements are not (or should not be) the modus
> operandi for a historian.

I'm not currently trying to write an article with citations, I'm saying
I've done a lot of research here and this is what I remember off the top
of my head, away from my references and not spending time to look things
up for a simple message thread.

Historian isn't my day job. Programming is, and has been for years. (I
was offered a book contract once for a history of Linux, but didn't have
time.)

(P.S. Citing wikipedia to back anything up is hilarious. It's pretty
much the modern definition of "non-authoritative reference". A secondary
source at best.)