----- Original Message -----
From: "Henrik Nordstrom" <hno@hem.passagen.se>
To: "Robert Collins" <robert.collins@itdomain.com.au>
Cc: <squid-dev@squid-cache.org>
Sent: Sunday, May 13, 2001 8:07 AM
Subject: Re: [squid-users] squid.conf: continuation lines?
> Robert Collins wrote:
>
> > I'll create a parser type for that then. IMO the existing acl file
> > inclusion method is quite useful - because the contents of the file
> > don't need the acl foo bar prefix for each line. This makes script
> > generated files easier. So I'd like to keep that method around. If
> > someone wants to suggest a more intuitive syntax for triggering it
(ie
> > acl foo bar source = /path/to/file) or whatever, that's ok by me
too,
> > and I'll endeavour to implement that.
>
> It is useful indeed, but the current config file syntax sucks and
should
> be changed to make it more obvious it is a file based acl definition.
> Possibly with a syntax somewhere along the lines you proposed above,
and
> most likely extended to dynamically reload the ACL on changes (maybe
as
> an option).
Cool. The config parser I've mostly built is all refcounted. (Built as
cbdata now, once I've propogated the changes everywhere I'll get dump
going again and then write a proper refcounting equivalent to cbdata. At
that point reconfigure should no longer be broken and I'll look at
asking for HEAD inclusion of some of the branch.
The point about refcounted configuration is that dynamic configuration
changes should become ___much___ easier. (It's as easy as removing the
config node reference, and then parsing anew. If someone else is squid
has a reference to the old node, they keep their config until they've
finished with it (ie an ACL) or are a module told to restart helpers or
the like.
However I have _no_ idea how to implement file watching in unix (doesn't
it require a signal?) at all, let alone portably :] We could walk the
config tree for file nodes every minute or two and re-read that node
though.
So I'll get it to the point that re-reading a portion of the config is
trivial, and then document it.
Rob
> --
> Henrik
>
Received on Sat May 12 2001 - 16:59:29 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:01 MST