[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Support character classes in glob authz rules

From: Julian Foad <julianfoad_at_apache.org>
Date: Mon, 03 Dec 2018 08:15:24 +0000

Branko Čibej wrote:
> https://issues.apache.org/jira/browse/SVN-4204
> https://issues.apache.org/jira/browse/SVN-4795
> Even though apr_fnmatch(), which we use for matching glob patterns in
> authz rules, supports character classes ([A-Z] etc.), we can't use them
> in the glob rules because of the way the config parser works. For
> example, the following rule:
> [:glob:/**/*.[Dd]oc]
> will be silently parsed to ":glob:/**/*.[Dd" [because] our config parser
> strictly follows the semantics of Python's ConfigParser and treats the
> first ']' it finds as the section terminator, ignoring any remaining
> characters to the end of the line.
> The proposed patch changes this behaviour as follows:
> * the /last/ ']' in the line is the section terminator;
> * only whitespace is allowed after the terminator to the end of the line.

Glad to see a proposal.

It makes me uncomfortable to depart from standard parsing. What if users are relying on Python ConfigParser or other compatible parsing as part of their Subversion authz infrastructure?

First I wondered if anything bad could happen if there's a silent change in meaning where a user has written, let's say,

> [:glob:/**/secret1] # was [:glob:/**/secret2]

It's hard to find any plausible example that would successfully parse and actually match something, but may be possible.

> The proposed change in the parser is only enabled for parsing authz and
> global group files, other Subversion configuration files will use the
> current semantics.

These sorts of quirks tend to make a system hard to maintain in the long run. What if we find another case where we'd like to depart from that parsing? How far would we go?

What alternative solution could we consider?

- Julian
Received on 2018-12-03 09:15:33 CET

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.