[whatwg] Fakepath revisited

[Aryeh Gregor]

On Sun, Sep 13, 2009 at 6:23 PM, Eduard Pascual <herenvardo at gmail.com> wrote:
> I already posted an example showing how fakepath can easily break
> compatibility with well-written sites. I explicitly asked for
> counter-arguments to it and none has been provided, but the argument
> doesn't seem to be taken in consideration at all.

I don't understand what the incompatibility would be.  You argued that
it would be a pain for some existing sites, but not how it would break
any existing sites.  Existing sites already need to strip off leading
paths, since all browsers except very recent ones provide leading
paths in this DOM attribute.  So what *existing* site could possibly
break here?  You seem to be saying something about filenames
containing "\", but since all existing sites must already strip out
the path, those would already break.

(Not that anyone uses files with such insane names in real life.  Try
running find / -name '*\*' on your local Unix system and tell me if it
returns anything.  I get nothing on my Linux desktop.)

> Hence I'm wondering how the compatibility arguments are treated here.
> Is compatibility with an unknown-size niche of clearly bad-designed
> sites more important than with potentially thousands of well-designed
> ones?

No, if you can demonstrate an actual compatibility problem in
practice, rather than a hypothetical issue that doesn't even appear to
be compatibility-related (AFAICT).

> Opera has claimed that they are keeping fakepath just because
> Microsoft claims some sites need it.

IIRC, Opera has some direct evidence that certain pages break.  I
don't think they're just taking Microsoft's word for it.

> However this group is willing to bend a standard based only
> on the claims from a single vendor... not to mention that this is
> precissely the vendor that less commitement has shown over the last
> decade on the area of web standards implementation.

HTML 5 seeks to be a specification that all major vendors are willing
to implement.  Microsoft is the most important here, since it's the
biggest vendor.  Its actions and attitudes toward standards are
irrelevant -- we're trying to build a useful standard, not wage a war.