W3C home > Mailing lists > Public > public-webapi@w3.org > April 2006

Re: ISSUE-47: responseText is not always accessible when readyState < 3

From: Anne van Kesteren <annevk@opera.com>
Date: Mon, 24 Apr 2006 12:58:47 +0200
To: "Ian Davis" <ian.davis@talis.com>
Cc: "Web APIs WG (public)" <public-webapi@w3.org>
Message-ID: <op.s8id39sx64w2qv@id-c0020.lan>

On Mon, 20 Mar 2006 16:14:38 +0100, Ian Davis <ian.davis@talis.com> wrote:
>> And what exactly do I have to implement? I think the specification  
>> should remain unchanged as what is currently specified is a sane thing  
>> to do. Unless of course there is some interoperability between the  
>> majority of browsers that is different from that what is specified.
> You're right, it's the sane thing, but it doesn't reflect reality. One  
> major browser throws an exception where we say it shouldn't. That means  
> that content authors can't rely on the behaviour we're specifying which  
> I believe lowers the utility of the text.

If we want down that road we might as well specify nothing, imho. The  
features we specify should be testable and it should be clear how they  
should be implemented. I think that authors reading specifications are  
well aware that what's in there is not always reflected by implementations.

>>  The above text could be used for the authoring guidelines though, I  
>> guess. It seems important to tell people you can't currently rely on  
>> what responseText its getter will do when readyState is not 3 or 4.
> I'd rather we wrote down what is reliable now and marked  
> incompatibilities as likely to be tightened up in the next version. I  
> think it's better to have a smaller spec that is accurate and  released  
> early.

Leaving all kinds of things undefined doesn't make a specification more  
accurate and I doubt it makes the specification any smaller. As an author  
I would much rather read a tutorial anyway, especially when there are  
differences between implementations. Things like "This method, when  
invoked without arguments, SHOULD do the same as invoked with its only  
argument being null" don't really tell me, as an author, that  
implementations might do otherwise. I don't think I'd really be able to  
distinguish SHOULD and MUST and on top of that see what kind of  
implications they might have on implementations...

Therefore I leave this text as is.

Anne van Kesteren
Received on Monday, 24 April 2006 10:58:55 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:16:21 UTC