kellermg at potsdam.edu
Mon Jun 14 20:00:48 GMT 2004
I started to write an SMB protocol plugin for Apache2. Then I got
distracted by a paying project and haven't been back at it. I spent many
hours planning out the requirements and it all looked very very doable.
My design had a separate "nmbd"-esque daemon just to deal with the
persistent LAN netbios-naming crap, but the "smbd"-esque part appeared
to be 100-160 manhours based on my Gantt. Of course, I could've grossly
underestimated it, but I don't think so. I'll probably hack back at it
sometime this summer. I really like the Apache2 infrastructure, and
after doing a couple toy protocols, wanted to jump in over my head. :)
On Thu, 2004-06-03 at 17:12, Luke Kenneth Casson Leighton wrote:
> btw if anyone is concerned about the speed of a multi-processed
> arrangement (one daemon per service, where services equals
> double-digits as in samba-tng) then of course all you need
> to do to fix the fork() issue is to use the same approach as
> apache(1&2): have a pool of pre-forked processes on tap.
> in fact i am surprised that 1) smbd has not had such code added
> to it already 2) that serious consideration has not been given
> to using the new apache2 framework: careful analysis of libapr
> shows significant code duplication and similarity in key areas.
> tng-technical at samba-tng dot org
"No one ever says, 'I can't read that ASCII E-mail you sent me.'"
More information about the samba-technical