[gclist] Web hosts for the FAQ

[David Chase]

> From: mps@geode.geodesic.com (Mike Spertus)
> I'm not against the overview of the technology, although I think it would
> be hard to do better than Paul Wilson's paper.

I think the aim was to hit the high points in a few pages.  That 
was my aim, anyway.

> ... However, the people who
> are considering whether to use GC are often most in need of getting the
> myths addressed.  Esp. when GC is coming more into the public eye like it
> is now through the growth of Java, Smalltalk, The actions in the ANSI C++
> committee and Bjarne Stroustrup's statements are also raising the profile
> of GC. The commercialization of C/C++ GC such as ours including advertising 
> and product reviews contributes to this as well. 
> 
> If the people who come to us looking for C++ GC are in any way representative
> of the people who would look at the FAQ, I can guarantee that many more of
> them want to know the answer to "Doesn't garbage collection just cover up
> your errors rather than fix them?" than to "How does incremental collection
> work?". In fact, when we promote C/C++ GC, people who call us with questions
> about GC ask basic myth-oriented questions many times more frequently than
> they ask technical questions (although they do ask both). I have no reason
> to believe that our experience would not apply to the readership of the FAQ.

I'm inclined to agree with this, but I don't think that I'm the person 
to write this.  My experience is that I've been using and porting 
the Boehm-Weiser collector since 1988, and I've never been disappointed, 
whereas I've seen products delayed and features dropped because of 
bugs in manual memory management.  Problem is, I have to keep reminding 
myself that consumers of computer technology have been "lied to" 
(to a first approximation) so many times that there's no particular 
reason they should believe me when I tell them that GC is the greatest 
thing since sliced bread.  I also have this problem speaking in marketese 
(i.e., not the whole truth) since I also know that strictly speaking, 
there is this very-low-but-not-zero-probability problem with optimizing 
compilers and conservative GC.  Thus, as an honest person, I cannot 
tell them there are "no problems"; I can merely tell them that they 
should worry more about their computer catching fire (1 incident) 
or their hardware silently doing wrong things (2 incidents, neither 
of them the Pentium), or bugs in their OS emulation of trapped instructions 
(2 incidents -- let's not get started on flaky network file protocols, 
or plain old hangs and crashes), or, gosh-golly, a compiler bug (countless 
incidents, including at least 2 that I myself accidentally put into 
a compiler in wide use in recent years). Anyhow, you get the picture.  
I'm not the right guy to write this stuff. 

I've been putting a bit of time each day into the FAQ the last few 
days -- people who expect to have opinions should give it a look 
at tell me what they thing of the content and the form.  It's picked 
up some content, been slightly reorganized, had a table of contents
and a lot of formatting added.

See  http://www.centerline.com/people/chase/GC/FAQ.html

David Chase