Ian Bicking wrote:
On Mon, 2003-01-27 at 12:03, Stuart Donaldson wrote:
  
Having a couple of examples in the documentation is good.
Or possibly having named alternates which could be invoked by a servlet 
in its own writeDoctype() method could ba a way to go.
Perhaps having a configuration setting would be good.
    

I strongly dislike having a configuration option.  Maybe that's because
I'm anti-configuration, but I don't think a configuration option gives
the right impression.

If Roger had a problem, then leaving 0.8's behavior like 0.7's would
have kept it from being a problem (I imagine).  If people are getting
wacky results because of the doctype, they still could get wacky results
from not having a doctype (just a different set of problems).

Page could use some documentation, but I think it should be very easy to
track down the doctype if that's a problem (though do browsers show the
doctype with view source?).  I would support proving in a docstring or
comment a variety of reasonable alternatives -- probably HTML 3.0, HTML
4.0/strict, and XHTML 1.0 (strict and transitional).  That seems
entirely sufficient to me.

What doctype was causing the problem?  Obviously a strict doctype would
be inappropriate, but were there problems with 4.0 transitional?

  
Of the potential options I put forward, the configuration option is admittedly my least favorite.

Since browsers use the DocType to determine how they will render information sent in the remaining HTML,  and since the remaining HTML is actually generated by Servlets and PSP code added on by users of Webware, it seems they should be the ones to choose what DocType to use.

I think that Page.py doesn't know what type of HTML will be output, or what type of rendering is desired so should probably not generate a DocType.  However since a DocType is required, sending the 4.0 transitional would seem to make sense as a default.

Chuck, the CVS log looks like you're the one that commented out the writeDocType() can you chime in with an opinion on how to deal with this?

It sounds like the current 0.8b2 behavior of not presenting a doctype is breaking 0.7 installations.  Should we consider going back to the 0.7 behavior and just make sure we have it documented well?  Should we leave it with the existing behavior?

If we are going to change anything, it would be best to get it into 0.8b2 which I would like to get out soon.

-Stuart-