From: Robin Rosenberg <robin.rosenberg.lists@dewire.com>
Date: Mon, 13 Oct 2008 17:10:50 GMT
Subject: Re: [JGIT PATCH 4/4] Intelligent parsing of ambiguously encoded meta data.
Message-ID: <200810131910.53891.robin.rosenberg.lists@dewire.com>
In-Reply-To: <20081013022708.GJ4856@spearce.org>

måndagen den 13 oktober 2008 04.27.08 skrev Shawn O. Pearce:
> Robin Rosenberg <robin.rosenberg@dewire.com> wrote:
> > We cannot trust meta data to be encoded in any particular way, so we try
> > different encodings. First we try UTF-8, which is the only sane encoding
> > for non-local data, even when used in regions where eight bit legacy
> > encodings are common. The chance of mistakenly parsing non-UTF-8 data
> > as valid UTF-8 is varies from extremely low (western encodings) to low
> > for most other encodings. If the data does not look like UTF-8, we try the
> > suggested encoding. If that fails we try the user locale and finally, if
> > that fails we try ISO-8859-1, which cannot fail.
> 
> Hmm.  I'm concerned about the infinite loop you have here.
> If ISO-8859-1 fails we'd be stuck here until the end of time.
> Plus its a bit ugly to read.
> 
> I wonder if this is any better.  It passes your tests and is 2
> lines shorter.

Yes. Not sure what I was thinking with the loop there... :)

-- robin