[E-Lang] language popularity

Mark S. Miller markm@caplet.com
Sat, 07 Jul 2001 16:11:59 -0700


At 12:08 PM Saturday 7/7/01, Richard Uhtenwoldt wrote:
>The thesis is that for a new language to become popular in today's world, it
>must be backed by a gorilla-sized organization (IBM backed PL/1, DOD backed
>Ada, MS backed VB, Sun/Oracle/Netscape backed Java, and they only succeeded
>because they rode the crest of the Internet wave and maybe also because IBM
>soon got on board in a big way) or it must have a fully-featured open-source
>implementation.

Yes, that's a good summary.  My only quibble is with the categorization of 
Java.  If Java had succeeded with an honestly proprietary approach, then, as 
you point out, this would still be consistent with the thesis.  However, as 
I've previously mentioned, and as the founders of Electric Communities can 
painfully attest, during Java's crucial early adoption phase, Sun was 
vigorously claiming that Java would be made open.  Gosling's statement 
that (approximately) "If Sun doesn't open Java, I will resign" was often 
repeated.  It was explained that Gosling didn't want to repeat the horror of 
NeWS (his postscript window system).  

Adoption of Java at EC was based on this claim.  Once EC found out the 
truth, of course it was too late.  While I don't wish to claim anyone at Sun 
knowingly lied, the net effect was the classic bait and switch.

"Fool me once, shame on you.  Fool me twice, shame on me."

All this happened before the term "open source", and before there was a 
clear demarcation between open vs proprietary licenses.  A few years later, 
after the term was coined and given a clear demarcation at 
www.opensource.org, I was at an amazing meeting between Bill Joy, others 
from Sun, Linus Torvalds, Eric Raymond, Brian Behlendorf, Larry Wall, Tim 
O'Reilley, and a few more I'm sure I'm forgetting.  At this meeting, Joy 
unveiled a prototype of Jini, and a draft of the Jini license, which was to 
become the SCSL license.  Joy wanted an endorsement in some form, by some of 
these people, of this license, to make it acceptable to the community.

Early in the meeting, when the terms of the license became clear, Linus 
stormed out of the room in disgust.  Although it took the rest of us longer 
to achieve the same level of clarity (or, perhaps we just preferred to argue 
rather than disengage?), none of us were ever taken in.  Why were we so much 
less vulnerable than EC had been in the past?  I think, because now there 
was a clear, public, consensus definition of "open" that had enough force 
that Joy, though he clearly desperately wanted to, never once claimed that 
SCSL was an "open source license".  And if he had so claimed, this claim 
could have been easily shot down.

AFAIK, Joy never got the endorsement from any of these people.


>C++ now has a full open-source implementation, but I think its userbase grew to
>almost-popular size before the open-source implementation was fully
>featured.  it was backed by ATT, which was at the time the largest single
>user of systems programmers, but had oddly little influence on software
>developers  outside its own walls.  can anyone say more about C++'s history?

I seem to recall the source to cfront from AT&T being generally available.  
While I don't know what the license terms actually were back then, people 
also paid less attention to these terms.  If source was available, then 
fixes were made and distributed without assuming that permission was needed. 
I believe we fixed some bugs in cfront during usage at Xanadu.


>judging by traffic on the comp.lang.* newsgroups, VB, C, C++, Java, Perl and
>maybe Python and Unix shell were the only massively popular languages 3 or 4
>years ago. [...and much more in the same vein...]

This is extremely valuable market research!  Have you (or anyone else) 
gathered this data together anywhere?  Is there a URL?  It would be great to 
cross-check this data with Zooko's SourceForge statistics and my informal 
sampling of shelf space at technical bookstores.


        Cheers,
        --MarkM