On 10/24/06 2:08 PM, "John Peacock" <[EMAIL PROTECTED]> wrote:

> [WARNING - speaking only for myself and not for the project]
> 
> Peter Eisch wrote:
>> Was there a reason the loadcheck plug-in (March 2006) didn't get included in
>> 0.32?  What are the requirements for predicting what plug-ins are in a
>> rollup?
> 
> As a rule, I won't automatically add a plugin to the repository unless
> it is something *I* want to run (and possibly support).  If there seems
> to be a very useful plugin that I won't/can't run (e.g. something to do
> with exim or postfix), then I will only add it to the repository if
> there seems to be a groundswell of support for the addition.  The wiki
> is that way ===> for everything else.
> 

I see you as a gateway to the management of the project.  If the project and
your participation as a manager exists to only support you as a user, then
you're not looking out for the project.

> We started to create a contrib folder as a place to centrally store
> outside developer supported plugins.  I haven't checked in everything I
> wanted to (see screed below about the refusal of ISO to add 10 hours to
> the normal night/day cycle).
> 
> In the above case, after finding the thread that included that plugin
> (inline, but see below), I think I probably didn't bother because it
> isn't all that useful (no offense).  If your system load is 7 or higher,
> you have a lot more problems brewing than worrying about e-mail being
> delayed.
> 

Where was the feed back, "I don't like your default value of '7' so I
adjusted the value to be '1' as a default.  Committed." ?

Your opinion about how an 8-way system shows load is your opinion.  It is
also a configurable parameter left to the user to tune.

As a project maintainer, you're not asked to necessarily have all the
answers, only to be available for reviewing, suggesting and supporting the
project.  If you refuse a patch, you owe the community the minimum of an
explanation -- especially since the one you cited, loadcheck, received some
reasonable accolades on the list last March.  Dare I say it was a
groundswell of support?

> I have a distinct corporate bias, where physical servers are cheap (but
> the rackspace may be the actual limiting factor).  All of our incoming
> e-mail is handled by extremely low-powered Cobalt RaQ servers
> (equivalent to P400's) and the only time I see loads above 1 is when I
> stupidly forget to tell all of the servers about a new domain
> (effectively mailbombing myself).  And even then, qmail is so damned
> efficient I can rack up multi-megabyte mail loops without so much as
> breaking a load of 4.  That's including scanning with two different
> virus scanners, BTW.

I don't follow what your point is, sorry.

> 
>> I recall my logging plug-in patch never got reviewed (Nov 2004).  I'm happy
>> to see it in current code -- I even see comments in the current code akin to
>> issues that I had.  It's cute and I'm mostly over it.
> 
> One convention that most lists have is that patches are accompanied by
> [PATCH] in the subject line (even if the patch is submitted as part of a
> larger thread, most mail clients will cope).  In several cases
> (including the ones you mention here), you have included a patch without
> a distinctive subject, plus when they are attached (and not inline),
> they come through like this:
> 
> --B_3184488241_38414932
> Content-type: application/octet-stream; name="syslog.patch"
> Content-disposition: attachment
> Content-transfer-encoding: base64
> 
> LS0tIHFwc210cGQub3JpZy9saWIvUXBzbXRwZC9BdXRoLnBtICAgIDIwMDQtMDktMjMgMTE6
> MTQ6NTYuMDAwMDAwMDAwIC0wNTAwCisrKyBxcHNtdHBkL2xpYi9RcHNtdHBkL0F1dGgucG0g
> MjAwNC0xMS0yOCAxMToyMjowNC4wMDAwMDAwMDAgLTA2MDAKQEAgLTI2MCw3ICsyNjAsNyBA
> QCBzdWIgU0FTTCB7CiAgICAgCiAgICAgICAgICAgJHNlc3Npb24tPnJlc3BvbmQoMzM0LCBl
> NjQoIlVzZXJuYW1lOiIpKTsKICAgICAgICAgICAkdXNlciA9IGRlY29kZV9iYXNlNjQoPD4p
> OwotICAgICAgICAgICN3YXJuKCJEZWJ1ZzogVXNlcjogJyR1c2VyJyIpOworICAgICAgICAg
> ...
> 
> meaning I have to save the file and open it before I can even see what
> is included (extra steps make it less likely that I will make the effort).
> 

I have no idea what you're referring to here.  The syslog diff was short,
simple and in-line.  I have no idea what "syslog.patch" is as I created no
such file yesterday or today.  If this is from some time ago, then I
apologize for using pine.  It is fair to refuse patches in-line.  Still, it
is nice to be able to review a patch in-line so as not having to save it and
view it with an editor.  Regardless, I had no feedback.  Ignoring
contributors for whatever reason is a disservice to the project.

Would providing an URL to the patch be better?  This can allow a universal
interface to the source/patch.  Let's try!

>> I've run with my own branch of 0.28 with logging as a plug-in and such for
>> about two years and I'm not really sure why I came back to the current.  In
>> other projects submitted patches typically a thumbs-up/down response with
>> some reasonable critique.  Granted I don't generate a large number of
>> submissions, but I wonder how many other bits of goodness have also been
>> passively overlooked in this project.
> 
> If I had to, I could document all of the washing machines and chainsaws
> I'm currently juggling, but frankly I'm afraid to face that list.  I'm
> able to do at least some Qpsmtpd hacking on work hours, since we rely on
> it so much, but I've been swamped lately.  I know that the other core
> developers are similarly afflicted with ETOOMUCHTODO and ENOTENOUGHTIME.
> I'm sorry if that is a lame excuse, but its the only one I've got.
> 

Same here, time is scarce.  It's frustrating to have to diff a release to
understand what things have changed, which haven't and figure out what's
missing.  I've literally diff'd every plug-in in order to get to 0.32 and my
experiences are reflected in my recent email.

> You posted three patches today before your rant, one of which is an
> unreadable (by humans) attachment with no useful description in the
> e-mail.  I hope you aren't offended that I haven't immediately applied
> them to the branch, but I'm just getting back from Real Life(TM) events
> of an unpleasant nature (funeral) and I don't know when I'll get caught
> up.  That's not your fault, however.  I'm just asking you to understand
> that things often fall through the cracks when part-time developers are
> involved.  Perhaps we should point people at an RT queue rather than
> relying so much on the list...

Where is there an RT system for qpsmtpd?  I don't know anything about any
ticket system.  I do know of the wiki.  This is typically a location where
things are documented or described, not a place where implementations are
considered or evaluated.  The wiki would be great if people maintaining
plug-ins on the wiki could have some sort of assurance that they'd be in
future releases.

Sorry be being a nag.  This repeated experience frustrates the elegance I
enjoy in qpsmtpd for handling email.

peter



Reply via email to