"C. Michael Pilato" <cmpilato@collab.net> writes:
> Daniel Rall <dlr@finemaltcoding.com> writes:
> > On Sat, 24 Sep 2005, Ben Collins-Sussman wrote:
> > > * the mod_dav_svn 'high level' operational logging still needs a
> > > quoting system, because it always describes paths in single quotes.
> >
> > Ben, what do you imagine this quoting system would look like
> > (e.g. doubled-up single quotes like with SQL)?
>
> Well, it being an Apache module, what about URI encoding the tokens,
> and leaving them whitespace-delimited?
We debated for a while in IRC, with Garret Rooney and Paul Querna (and
a bit of Justi Erenkrantz toward the end), and concluded that for
human readability, quote-delimited non-URI-encoded paths would be
better than whitespace-delimited URI-encoded paths.
They're more or less equally easy to parse, but much of the time it'll
be a human reading these logs. People are already accustomed to
backslash-escaped quote-delimited strings. Obviously, they could
manage URI encoding too, it's just a question of which comes more
naturally. Also, there was the argument that conceptually, these
really are paths within the repository, not URLs.
Here's the IRC transcript:
<chipig> we url encode paths in logging.
<chipig> as thats what modules see.
<kfogel> chipig: is there any other delimiting?
<kfogel> or is it just whitespace (which would make sense to me, since
spaces in the actual paths will be encoded)
<chipig> whitespace in this context I think makes the most sense.
<chipig> but, svn users expect file system path
<chipig> which, are not url encoded.
<kfogel> yeah :-(
<chipig> (har har har... But, I would just do everything URL Encoded..)
<kfogel> We just have a clash of expectations here, it's really not
clear which is correct.
<chipig> most admins can figure out url encoding.
<kfogel> But, frankly, I think users are going to think of these as
paths, not URLs.
<kfogel> users == admins, in this case
<chipig> this is going into the httpd error log?
<rooneg> well, generally people will probably send it to its own log,
but it does go through the httpd logging stuff.
<chipig> how will they send it to its own log?
<kfogel> no, it's generally not going to error log
<kfogel> there's a svn logging conf directive, which tells apache
where to send it
<kfogel> our assumption (and our example) is to a dedicated svn
logging file, not to standard apache logs
<kfogel> It is, internally, using httpd logging mechanisms to achieve
this, of course.
<chipig> erm.
<chipig> svn is adding a SVNlog type directive?
<kfogel> chipig: I don't remember the details, but I can find them out
if you want.
<chipig> ErrorLog svn_log env=svn_log
<chipig> ErrorLog error_log env=!svn_log
<kfogel> chipig: btw, a complete list of the several revisions that
implemented this can be found in issue #2409
<chipig> somehting like that, is okay.
<kfogel> chipig: right
<kfogel> So, given that these logs will most often be custom log
locations, and be read with a path mentality rather than a URL
mentality, I think I still lean toward backslashes, as
intriguing as the URL-encoding idea is.
<chipig> oh, right. its actually a CustomLog
<kfogel> yup, I recall that now
<chipig> they are already single quoted in the suggested log format
<jerenkrantz> kfogel: ?
<chipig> CustomLog logs/svn_log "%t %u \"%{SVN-ACTION}e\"" env=SVN-ACTION
<chipig> actually, double quoted :)
<kfogel> jerenkrantz: we're discussing how to quote paths in the
custom mod_dav_svn operational logging
<chipig> it doesn't really matter, I guess.
<jerenkrantz> ah.
<kfogel> jerenkrantz: see backscroll, or I can supply with an edited
paste if you need, just let me know
<chipig> I would go for, whatever is easier to write.
<kfogel> chipig: they're equally easy, really. I want to go for what
will be best for users/admins.
<chipig> (in the case of URL encoded, its at worst one function call)
<kfogel> chipig: I think you're misreading that.
<kfogel> chipig: the overall action string may be double quoted in our
suggested example, but it's the paths *within* the action that
we're talking about here.
<chipig> yup.
<kfogel> chipig: it'll be one function call either way, the only
question is whether we have to implement the function or if
it's already implemented :-).
<chipig> i think its a tiny bikeshed. I think url encoding is best,
since it doesn't bother me to read it, but users expect
non-url-encoded, and this is meant for them, so, give it to them.
<chipig> it might make parsing it harder
* kfogel nods
<chipig> if tis non-url-encoded
<chipig> but, thats a tiny detail.
<kfogel> not much harder though
<kfogel> yeah
<kfogel> these are most often to be human-read, I think
<chipig> harder to use awk :)
<kfogel> actually the parsing is a bikeshed too, they're equally easy
in the end
--> brane has joined #svn-dev
--- ChanServ gives channel operator status to brane
<kfogel> okay, so we're agreed: green.bikeshed.com, right? :-)
<kfogel> Ah, the illustrious brane!
* rooneg wants the bikeshed to be blue
<brane> i prefer green to illustrioius
<chipig> http://blue.bikeshed.com/
<jerenkrantz> kfogel: sounds like chipig helped you. anything more i
can offer?
<kfogel> jerenkrantz: I could use an oil change for my car, but I
think as far as logging goes we're set.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Sep 26 20:36:19 2005