Problem with nginx/0.8.54 and alias

[Maxim Dounin]

Hello!

On Fri, Feb 18, 2011 at 12:46:44PM -0500, Jim Ohlstein wrote:

> On 2/18/11 11:57 AM, Dayo wrote:
> >Здравствуйте!
> >
> >Maxim Dounin Wrote:
> >-------------------------------------------------------
> >>>>Just a side note: forum is awful, it eats
> >>data.
> >>>
> >>>To be fair to the forum, I was the one that
> >>truncated the quote to only
> >>>show the relevant part to my post.
> >>>The long and complex quotes become impossible
> >>for me to follow.
> >>
> >>I'm complaining about "" in regexp being eaten by
> >>forum engine,
> >>nothing else.  Forum's inability to preserve data
> >>intact is really
> >>distressing.
> 
> I am unaware of this particular issue until now, or, if I was it has
> slipped my mind. I have responded to your complaints in the past and
> fixed as many of them as possible. Instead of making these snide
> comments on the mailing list, it would be more helpful if you simply
> aired your issues with me privately. I then *might* be more inclined
> to work on them. It would be better to handle forum issues on the
> forum (or privately) and not to pollute the mailing list with them.

Sorry for being rude, but I've noticed the issue and commented 
according to it's nature.  I indeed consider anything which causes 
data corruption to be awful (and this isn't the first data 
corruption issue with forum, unfortunately).  I also consider 
nginx proxy cache to be awful for similar reasons - hope this will 
help you to feel yourself better. ;)

And as far as this indeed affects mailing list and forum, 
including people in this particular thread (note several messages 
with incorrect configs due to <NAME> being stripped) - I don't 
consider private reporting to be better way to go.

Some more issues in additional to one with <NAME>, in case you are 
willing to work on them:

1. BB-codes seems to be processed in mailing list messages shown 
in forum, thus providing another data damage vector.  Example may 
be seen in the same thread, in Igor's message:

http://forum.nginx.org/read.php?2,176110,176123#msg-176123

2. Word wrapping in UTF-8 seems to be done based on bytes, not on 
chars.  This causes really narrow (and hard to read) messages from 
forum in russian mailing list.

3. Subject isn't correctly encoded - UTF-8 chars may be splitted 
into different MIME words (which is not allowed per standard).  
Igor currently workaround this with local script to re-encode them 
correctly, but this in turn may damage other Subject's.  Original 
value is preserved in X-Subject header, here is some recent 
example:

X-Subject: =?UTF-8?Q?403=20Forbidden=20=D0=BF=D1=80=D0=B8=20=D0=B2=D1=85=D0=BE?=
        =?UTF-8?Q?=D0=B4=D0=B5=20=D0=BD=D0=B0=20=D0=B2=D1=82=D0=BE=D1=80=D0?=
        =?UTF-8?Q?=BE=D0=B9=20=D0=B4=D0=BE=D0=BC=D0=B5=D0=BD(?=

Note "=D0" and "=BE" (single UTF-8 char) split across MIME words.  
It would be good to resolve the core issue with forum and remove 
the workaround in question.

Maxim Dounin