Re: ajax bookmark problem

Tony Rice wrote:

> I"d like to hear critiques on the following method for dealing with the
> back button and bookmarkability problem with AJAX. Whenever I do
> something on a page with ajax, I add to document.location.hash (which
> doesn't reload the page but does include the parameters in a bookmark)
> like so:
>
> document.location.hash = document.location.hash + '&param=value';
>
> So the URL gets updated like so:
> http://host.com/cgi-bin/foo.cgi?cmd=bla#&param=value;
 
Not really.

Zeroth, `document.location' is deprecated long since. Use `window.location'
instead.

First, you add to location.hash. However, location.hash may or may not
contain the leading `#' character, depending on the UA's DOM.

Second, there are user agents which do not honor RFC3986 in this regard,
meaning that a "&foo" is considered a part of the URI's `query' component
even if it is after the `#'. You should therefore omit the `&'.

Third, you have to escape (percent-encode) the content of the query-part or
the hash component.
 
> When that page is loaded from that bookmark, I'm converting that
> information in the hash to the query (which reloads the page with those
> parameters in the URL and drops the hash):
>
> var hash = document.location.hash.substr(2);
> if (hash.length > 0){
> var searchadd = '';
> var elements = hash.split('&');
> for (i=0;i<elements.length;i++){

`i' is undeclared, therefore either it is global or the assignment can
result in an error. Comparing against the unchanged elements.length in
every loop is inefficient.

         for (var i = 0, len = elements.length; i < len; i++)
         {

> searchadd = searchadd + '&' + elements[i];
> }
> document.location.href=document.location.pathname +
> document.location.search + searchadd;

However, this does not make much sense. First you split the fragment part,
then you concatenate it again.

You should choose another delimiter for the fragment part components, then
use `&' to join them and concatenate with location.search. Example:

  window.location = window.location.pathname + window.location.search +
    "&" + hash.split("_").join("&");

> }
>
> Is it correct that the hash (i.e. everything after the # in the URL) is
> only available client side?

It appears so. Even with PathInfo, I have not been able to create a
request that includes the fragment component. However, I remember to
have read something about URL rewriting which used the `#' in a request;
unfortunately, I do not remember where.
 
> The reload to incorporate those parameters in the URL is time consuming.
> It takes an additional 60-90 seconds over a URL that doesn't need to be
> corrected. Is there a better way?

Efficient programming.

HTH

PointedEars
Received on Tue Feb 7 21:29:16 2006