home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!swrinde!emory!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!andrew.cmu.edu!pa0q+
- From: pa0q+@andrew.cmu.edu (Phil Andrews)
- Newsgroups: comp.graphics
- Subject: Re: SDSC Image Tools source
- Message-ID: <kfHQTts000009CQ0UC@andrew.cmu.edu>
- Date: 8 Jan 93 08:29:13 GMT
- Article-I.D.: andrew.kfHQTts000009CQ0UC
- References: <1ifpmsINN33k@network.ucsd.edu>
- Distribution: world
- Organization: Pittsburgh Supercomputing Center, Carnegie Mellon, Pittsburgh, PA
- Lines: 58
- In-Reply-To: <1ifpmsINN33k@network.ucsd.edu>
-
-
- In response to the SDSC Image Tools release posting, Alaxander-James Annala
- wrote:
-
- >I don't know what experience with other supercomputer centers may
- >be -- however, the gnu people don't have trouble with supporting
- >'modified' copies of their software. From a user's perspective,
- >these binary only distributions can create problems -- for example
-
- As the graphics coordinator of the Pittsburgh Supercomputing Center I can
- add our ten cents worth. Many people seem to think that writing a useful
- utility is the tricky part. This is just not true. Almost by definition
- the effort
- involved in any successfull system will be dominated by support, and efforts to
- reduce the support requirements of a freely available system are the most
- important part of its development.
-
- Once you decide to try for some reasonable amount of bug fixing the most
- difficult scenario to deal with is the unauthorized user who makes changes to
- the source code which (perhaps unkown to him) introduces new bugs. Trying
- to deal with these, knowingly or unknowingly, is a frustrating experience
- guarranteed to pursuade most people never to release anything on the net
- ever again. SDSC deals with this, perfectly reasonably, by not releasing
- source.
- This is cannot be all peaches and cream for them however, because it means
- they must maintain a very large number of binaries and the
- machines/environments
- to produce them.
-
- In our case the PSC has been distributing several packages for over five years
- now, our biggest "seller" being the CGM interpreter GPLOT, via anonymous FTP
- (don't send me mail, it's on ftp.psc.edu). We worried about this problem
- from the
- start and decided to approach it by making the source freely available but
- copyrighting it to explicitly prevent the redistribution of changed
- sources/binaries.
- And yes, we still have had people do this, and I have sent them nasty messages
- telling them to stop. We also provide as many binaries as possible.
-
- Other centers have other approaches; the software tools group at NCSA
- explicitly
- places their code in the public domain (please correct me if anyone wants to),
- and it is still not clear to me whether there even is a "best" approach.
- In defense
- of SDSC, we have noticed a clear trend towards more of our users taking
- binaries
- directly rather than wishing to deal with the sources and I expect this
- trend to
- continue as compilers, etc., become unbundled. However, I can believe that
- the availability of source gives even these users a warm fuzzy feeling, simliar
- to a decent warranty on a piece of electronics.
-
- One thing that is definitely true, if you worry too much about 100% of users
- being able deal with your source, you are locked into arcane and unproductive
- programming practices and out of advances such as ANSI C and C++, let alone
- more aggressive approaches such as SmallTalk and Common LISP.
-
- -Phil Andrews
-