httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From r..@ai.mit.edu (Robert S. Thau)
Subject Re: project plan
Date Tue, 16 Jul 1996 00:43:09 GMT
  I guess a pertinent question might be, to what extent does rst's threading
  package lock us into user-level threads?

Only to the extent that we are committed to its interface, which is
pthreads-like for the most part, but not exactly the same.  It would
be possible to design a compatibility veneer which hid either rsthreads
or a pthreads interface, depending on whether pthreads were available.

In any case, there are reasons for supporting threads of some form on
platforms without kernel threads --- which are, I will remind everyone
yet again, support for session layers and allowing modules to filter
the output of other modules.  These essentially require some sort of
threading machinery (user-mode perhaps) even on legacy systems, unless
we want to simply abandon support for advanced features on those platforms.

FWIW, rsthreads is already the most widely-ported of the numerous
user-mode threads packages I've looked at in the course of doing this
stuff, and the easiest to port (typically requiring less than ten
lines of new non-portable code); in fact, the whole package is
comparable in size to the support for *two machines* in Chris
Provenzano's user-mode pthreads (though that is admittedly an extreme
case; most require only a couple pages of assembler per machine).

rst

Mime
View raw message