Don't break existing links!

[Michael Nahrath]

George Pauliuc <pauliuc@gmx.net> schrieb am 2002-12-06 18:04 Uhr:

> On Vi, 2002-12-06 at 17:38, Michael Nahrath wrote:
>> There is no special announcement to the user neccesary.
>> Sending http: error code 301 _is_ the announcement to the UserAgent.
> 
> The UA means in common terms browser.  And browsers are known not to use
> all standards and to badly implement the standars they support.  Are you
> sure this is a valid option?

Quite shure. 

Means: "I have tested with some dozends of differen browsers on several
platforms over the last three years repeatedly and have never seen any
problems with it." 

I have postet the link to my own homepage as an example. I guess I'd know if
it wouldn't work reliably.

My server's logfiles showed that there was one defect spider in the end of
2000  that didn't get it right. I have not seen it again since.
 
>> No need to bother the user with this. "Don't mention the mechanics!"
> 
> Telling the visitor in plain language "this is the link you should use
> from now on" has nothing to do with mecanics.  And avoids lots of
> trouble.

Telling him what to do _is_ trouble.

Giving him the content he wanted to get directly is what he expects and what
he should get.
 
>> Some scripting languages are able to send fully correct HTTP-headers too.
> 
> Wrong.  All scripting languages that have some connection with the web
> and can send headers CAN do that.

I can't see the contradiction in those statements ...

>> Maybe Browsers would expand the relative URL to a fully qualified one if
>> they get it via <meta http-equiv="..."> but I am not shure about this.
> 
> Plase, please, make up your mind.  You say "you can't" and three
> paragraphs below you say "I'm not sure".

OK, you need it more precicely:

<http://www.w3.org/TR/html401/struct/global.html#edef-META> says that
webservers may gather this information.

AFAIK no web server does scan each file it delivers for additional headers
provided by the 'meta' element.

Some browsers do interpret them on their own.
This is not standard behaviour and should not been relied upon. If they
expand relative paths upon this, it is even less standard behaviour.

Can you be shure that they will continue to behave (according to the words
of the specs: wrong) like this in the future? Who can?
 
>> IIUC negotiating <http://www.gnupg.org/documentation/index.html.de to
>> http://www.gnupg.org/(de)/documentation/index.html is a Caudium only
>> "feature". That will make portability to other servers a problem anyway.
> 
> So... let me think.
> 
> I say something about standards, code cleanliness (is there such a word
> in English?) and portability.  You answer with "is a Caudium only
> feature".  What is the logic of your answer?

Sorry, my mistake. I really should not use irony on international lists.

I shurely don't see this kind of URL rewriting as a positive "feature".

Instead I call it a "bug" as it breaks with common practice of dealing with
URL (since it comes to the user this _is_ a valid argument), makes
maintaining the site and linking to the pages harder and raises a problem
for portability. 

Greeting, Michi