W3C home > Mailing lists > Public > www-tag@w3.org > July 2002

RE: "URIs, Addressability, and the use of HTTP GET"

From: Joshua Allen <joshuaa@microsoft.com>
Date: Tue, 9 Jul 2002 13:53:41 -0700
Message-ID: <4F4182C71C1FDD4BA0937A7EB7B8B4C105BA2938@red-msg-08.redmond.corp.microsoft.com>
To: "Tim Bray" <tbray@textuality.com>, <www-tag@w3.org>

> > Regarding "Myth: URIs cannot be longer than 256 characters" -- I
suggest
> removing this block.  It seems to imply that URI is the same as a URL
with
> http: scheme, and talks about server limitations only.  Many uses of
URI
> do not require a server at all, and client limitations still exist
with
> respect to the narrow subset of URIs that serve as http: URLs.  In
> particular, mobile devices such as PDAs and cellphones, which are
widely
> deployed and used, have all sorts of differing restrictions on length
of
> http: URLs...
> 
> I'm surprised - I've been doing web stuff since '93, in many different
> kinds of scenarios, and have never seen any system failure traceable
to
> the URI being too long.
> 
> Could we maybe see some data?  Who knows about actual deployments out
> there in the wild that exhibit this kind of breakage? -Tim

http://www-106.ibm.com/developerworks/library/wi-entwap/?dwzone=wireless
mentions that many gateways need shorter URLs, but they don't suggest a
limit.
http://www.weblogic.com/docs51/classdocs/wap.html cites a potential
128-character limit in their docs.

The reason that most people don't put a hard number on the limit is
because the limits on cellphone URLs in practice vary wildly depending
on device, carrier, gateway vendor, and so on.  I have seen exhaustive
tables on the net where people have figured out and recorded the
limitations, since the vendors are rarely in a hurry to disclose their
limitations.  I don't have time to go through my bookmarks and find all
of the good sites.  One used to be at:
http://internetalchemy.org/wap/ualist.html

More at http://www.allnetdevices.com/faq/?pair=07.014

Trust me, I have spent many, many hours gnashing my teeth over these
issues, and cursing the people who told me I could blindly trust long
URLs to be supported.  These issues are nowhere close to being obsolete
in practice (lots and lots of 7110s out there still, and those are some
of the worst offenders).  Sometimes it is actually necessary to use a
POST instead of GET, because POST is supported by device, while the
gateway hacks URLs.  I'm not saying that real-world implementations
*break* because of these -- I am simply saying that people work around
the limitations, and this occupies a whole lot of time for the average
person serving up content or services to mobile and embedded devices.
They *often* have to deal with nasty URL length limitations, and
workaround them in ways that would make RESTians shudder.

I also think it is probable that this situation will prevail for a
number of years.  Cell phones may eventually evolve to the point where
it's not an issue, but then new devices will have moved in to be the
cutting edge of web connectivity.
Received on Tuesday, 9 July 2002 16:54:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:09 GMT