Re: lynx-dev Javascript

[David Woolley]

> 
> I think the last two are of the most demand by the lynx users
> since we cannot find a "submit" button otherwise on those pages.
> But this seems a little expesive to implement all javascript
> for this (estimated) lower frequency issue.

Where I've done submit processing I've revalidated the form first.
The way I've done it is Lynx safe, because I have the submit button
be a normal input type=submit, and then have an onSubmit action on 
the form, which revalidates and returns false if it doesn't like
something.  The false return causes a JS aware browser to abort the 
submit.

The reasons for revalidating are:

- if the original check is done on onChange events, it is possible
  to ignore the error messages;

- sometimes fields have to be validated in combination, or in a non-modal
  context, may depend on which submit button is pressed.

At least one reason for not allowing a non-JS submission is that you 
don't want to have invalid data submitted.  Another reason may be that
details of the submission may be determined by combining multiple
fields.

Even with code that does no graphic trickery, location setting, style
manipulation etc., I have had enough problems in producing code which 
is safe on both IE4.01 and NS4.5 without making user agent tests.  That
indicates to me that even just form submission can stretch the JS 
capabilities of the commercial browsers.