Re: <Form><input type="file" value !="browse" />?

> Maybe, but then they need to have the user agent tuned to use an image
> that they recognize. Using an image as a page-specific way of implementing

As I've said before, although Jon didn't understand it, and thought
I was joking, what he is really trying to do is to write his own user
agent, using HTML, CSS and browser (proprietory) automation features as
his user interface graphics library.  This is, to some extent, common
practice in the general commercial use of web browsers.

Whilst he fails to get this point, and thinks he is writing web pages,
I don't think he will really understand your point here.  He's also
trying to simulate things that really need a special markup language
or extensions to the existing ones.

I'd like to be able to say that the right way of writing your own user
agent would be to start from Mozilla sources, but compared with hacking
things with browser object models, there is just too steep a learning
curve.  It's this massive job of writing a custom user agent at this
level that is probably why he thinks I'm joking when I say he's writing
his own user agent, but, at least in principle, one could create a user
agent creation kit that made custom user agents easier that carrying
the user agent with the document (although many people would then
ask how to do things that are not possible because of the simplification,
as they do for hacking user agents with HTML etc.)

(Things tend to be worse with commercial use of user agents written
in HTML, as they usually want to force their user agent on every user
of the page, as part of their corporate idendity, exacerbating the 
problem I was arguing with Joe Clark about design leading to a lack
of consistency in user interfaces, making the web difficult to learn
from cold.)

Received on Wednesday, 12 March 2003 03:19:51 UTC