Re: signature/encryption tags -> follow-up

Philip Trauring (
Mon, 10 Apr 95 20:24:29 EDT

>What you are really proposing here is not really HTML, but some other
>media type, something like application/pgp-signed-document, or
>more appropriately, a transfer encoding of pgp-signed-document
>which the user agent automatically decodes according to its local
>rules (in .mailcap on unix).
>By using an encoding type, I can send a text/html or an image/gif and
>still use the same transfer and verification machanism.
>In other words: yes, pgp may very well be used to check signatures on
>HTML documents, but pgp has its own mechanisms for doing this - and
>HTML should not adopt them in favor of some other mechanism.

Your scenario does not allow multiple-segment files.

>A browser has to integrate crypto support if it is to take
>advantage of it - and if you are going to integrate support for crypto
>into the browser anyway, why use something as arbitrary as PGP's encoding?
>Why not actually use a more general mechanism?

It is a simple matter to call out to the external program to do the
calculations. By implemenating the form=???? comment in the <SIGN> tag you
can create definition files for the encryption programs so the HTML file
can use whatever format it wants.

>As someone else noted earlier, the signature role can be dealt with by
>expanding the use of the MD attribute to apply to all containers in
>HTML, so you could say
> <BODY MD="opaque-md5-hex-string">
> stuff that is digested goes <B>here</B>
> </BODY>
>To verify:
>compute the MD5 of the text between <BODY> and </BODY>, and compare with
>the stored value.

MD5 is not a signature algorithim. It is only a one-way hash and does not
provide authentication. In order to provide signatures you need encryption
algorithims such as the RSA ones used in PGP(the signature is basically an
encryption MD5 code).

>As to whether HTML should be laden down with a lot of cryptography - I
>think the answer is no. HTML is a markup language, not a keyring format
>or a cryptography application. Cryptography applications may manipulate
>HTML, but I think that is outside the scope of the HTML specification
>beyond the minimum tags needed to support them (i.e. the MD attribute and
>possibly a defined LINK name to refer to the public key of the author).

Integrating a signature tag is the only way to provide real document-level
security in WWW documents.

Philip Trauring

Philip Trauring
Brandeis University MB1001
P.O. Box 9110 "knowledge is my addiction,
Waltham, Ma 02254-9110 information is my drug."
(617) 736-5282 ['94/95]

WWW home page: