[Zope-dev] [Proposal] Allow (almost) arbitrary ids in Zope2

[Dieter Maurer]

= Introduction =
Zope2 allows only a very restricted set of characters in
`ObjectManager` ids -- a subset of the characters which can
be used unescaped in urls. Neither non-ASCII characters are allowed
nor some quite important ASCII characters.

When Zope is used outside the english language domain and
accessed via WebDAV, then the restriction to ASCII characters only
is heavily felt by WebDAV users, as they are accustomed to use
readable names (with
non-ASCII characters) in their file systems.
They cannot understand why such objects cannot
be transfered into Zope (especially given the most unhelpful error
messages of some WebDAV clients).

The limited set of allowed ASCII characters leads to other problems:
e.g. email addresses cannot be used as ids as `@` is disallowed.

= Feature =
Allow almost arbitrary characters in ids.

A few restrictions remain:

 * ids cannot start with a '_' because they would not be traversable

 * ids cannot contain '/' because Zope code uses
   `urllib.[un]quote` on complete urls (a bug, [un]quoting must
   only happen on the individual pieces, never the complete object)
   and `quote` by default interpretes `/` as safe (not to be quoted).
   Allowing '/' would risk too many breakages.

Allowing non-ASCII characters in ids poses a risk, as
the uri standard (RFC 2396) does not allow to specify the
encoding used in urls. There is only HTML 4 which recommends
that url components are first encoded as utf-8 and then url escaped
as necessary. But this recommendation seems not to be widely used.
However, when urls are constructed in a client (rather than directly
fetched from an HTML source or given by a user) e.g. in Java or !JavaScript,
then it is not unlikely that the recommendation is followed (as
these systems work internally with unicode and need to use some
encoding for url construction). I also have seen a client
(the MS WebDAV client) to interpret ("binary") urls in some
encoding and then recode them with utf-8.

I see two options:

 * Zope follows the HTML recommendation (and uses utf-8 in its generated urls).

   This implies that either ids need to be unicode or their encoding
   known (i.e. specifiable).

 * Zope does not follow the HTML recommendation and uses
   "binary" ids (with some unknown encoding) but uses some heuristics
   to handle HTML 4 compatibly generated urls.

As I am mainly concerned with WebDAV access and the WebDAV clients
I have seen so far use the binary id interpretation, I favor the
second option.

-- 
Dieter