On 7/19/07, Justin Erenkrantz <justin@erenkrantz.com> wrote:
> On 7/17/07, Erik Huelsmann <ehuels@gmail.com> wrote:
> > This problem could be solved to by adding the ICU lib as a dependency
> > and change all path comparisons to use the ICU normalform agnostic
> > comparison routines.
>
> I am pretty much against the idea of adding a *required* 39MB (~10MB
> gzip'd) library that isn't widely available as a requirement to run
> SVN. Yes, Debian/Ubuntu has it as do some others; but it's not that
> popular; quick look at the release notes indicates that ICU requires
> VC2003+ for compilation - so does it actually support VC6 (listed as
> "rarely tested")? We had enough Win32 people revolt against APR-iconv
> - now, we're going to backtrack and add in a *huge* library to replace
> it? Yikes. This is like squashing a fly with a sledgehammer, IMO.
>
> Why can't we just do input validation ourselves? This is clearly a
> very specific corner case and one we can detect quite trivially (i.e.
> look for the chars that only exist in normalization form D).
Input validation isn't enough. We can't guarantee that all paths
internally will be of the C form, if the admin isn't able to control
the clients setups. (Older Mac clients will still be able to submit D
form paths).
mbk said that there's no reason in supplying backward compat for macs,
since they are already broken, but the broken-ness only shows when
working in a mixed Windows/Mac/Linux setup.
Essentially, the problem shows up on Macs, but Subversion is broken
here, not the Mac.
bye,
Erik.
PS: The ICU library can be can be reduced in size to 1/10th if you
only want normalization.
PPS: Did people really revolt against apr-iconv because of its size?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Jul 19 08:54:02 2007