IE, Apache Clash on Web Standard

sbsea1 writes "Here is another instance where Microsoft is going one way and everybody else going to other. eWEEK Labs found that Microsoft is using a different implentation of digest authentication which differs from the W3C's digest authentication standards. Internet Explorer Version 5.0 and higher--as well as Microsoft's IIS Web server--has a significant security incompatibility with other major Web browsers and with the Apache Software Foundation's Apache HTTP Web server."

Comments

People, please read the article

[PhysicsGenius]

I hate Microsoft as much as the next Slashbot, but let's get the facts before we post frontpage Slashrants. It says, and I quote:

We [microsoft] were told by the Apache group that it would support multiple digest protocols. The MS Digestion protocol just hasn't been implemented by Apache yet.

In other words, like the libXML problem we all remember from last year, this is Apache's fault.

Differences in computing the message digest?

[cpeterso]

Does anyone have any real information about the actual differences between how Microsoft and Apache are computing the message digest? The article does not say much. I know the Microsoft and Netscape used to have some interop problems because one implemenation (Netscape's, I think) would include a string's NUL terminator when computing a message digest. This would obviusly lead to a different result. :)

who cares about digest authentication anyway?

[phr2]

Sending the digest in the clear still makes most user passwords vulnerable to offline dictionary search. Digest authentication was a kludge on top of HTTP basic authentication (which sends the naked password in the clear) designed at a time when SSL was scary and complicated and there were no free SSL web servers.

These days, for casual passwords like /. logins, HTTP basic authentication is still usually good enough. For passwords that need real security, use mod_ssl instead, which is easily added to Apache 1.3 and comes with Apache 2.0 by default, and do basic auth over SSL so the whole HTTP stream is encrypted including the password. HTTP digest authentication's security is sort of halfway between HTTP basic auth and HTTPS basic auth. As a halfway measure, it's not really that useful any more.

Re:who cares about digest authentication anyway?

[Zeinfeld]

Digest authentication was a kludge on top of HTTP basic authentication (which sends the naked password in the clear) designed at a time when SSL was scary and complicated and there were no free SSL web servers.

SSL did not exist when I invented the Digest mechanism. The problem was the patent on RSA and Diffie Helleman.

Digest was invented for one reason and one reason alone which was to provide a replacement for BASIC and avoid sending password in the clear.

Microsoft implemented Digest first, but Netscape refused. This was before they hired a credible security person. They believed that sending passwords over the internet en-clair was a less important security issue than protecting the authentication information in the Web server storage.

Microsoft removed Digest from IE in IE4 as Netscape refused to implement. Then the IETF stated that HTTP could not become a standard if it sent passwords en-clair at which point people pulled the draft out again.

Removing Digest from IE was not a big issue for me since if only Microsoft was going to implement the standard they might as well use the NT password authentication scheme.

The dictionary attack issue is important, but it was not possible to address it given the state of the IPR at the time. If Diffie Helleman had been available I would have designed the protocol entirely differently. It would have been possible to address security of the auth data on the wire and in storage.

For passwords that need real security, use mod_ssl [modssl.org] instead, which is easily added to Apache 1.3 and comes with Apache 2.0 by default, and do basic auth over SSL so the whole HTTP stream is encrypted including the password. HTTP digest authentication's security is sort of halfway between HTTP basic auth and HTTPS basic auth.

Actually I would recomend Digest over HTTPS. The problem with BASIC is that you have to trust the end point, that is fine if the application is such that the application justifies buying a certificate or securely distributing the point of trust.

More generally however I would suggest people look at our more recent work in SAML (security services at www.oasis-open.org)