Re: Thread implementations...

Dean Gaudet (dgaudet-list-linux-kernel@arctic.org)
Sat, 27 Jun 1998 12:47:57 -0700 (PDT)


On Fri, 26 Jun 1998, Linus Torvalds wrote:

> But as far as I'm concerned, the kernel is better at looking up filenames
> than apache will ever be unless you guys start really doing well.

Sure, except apache doesn't have to look up filenames, it needs to look up
URLs... and URLs don't always map to filenames (sometimes the mapping is
non-obvious, like it is for proxy servers). The flow engine I've been
playing with treats URLs as opaque objects. I can use whatever cheap
table lookup data structure fits the ticket... I don't need to break apart
path components, no access checks, those are all done once -- before the
object is put into the flow cache. The result of the lookup is an open FD
or a mmap region (plus metadata).

> Note! I don't think apache is going to be the best thing to use with this,
> if only because apache tries to be too clever. This is really meant for
> something that
>
> - uses threads

You mean "uses one kernel-task per connection". Processes are fine too.

> - does no caching _at_all_, because it knows the kernel can cache
> everything better than most user mode programs

You're so sure of yourself ;) I'll take that challenge.

At any rate, you describe apache 1.3. I'll work out a patch for you after
the weekend.

Dean

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu