1996-03-20 - RE: Microsoft’s “answer” to Java

Header Data

From: Blake Coverett <blake@bcdev.com>
To: “‘Alan Olsen’” <alano@teleport.com>
Message Hash: 805a8454ed7f598a0d64f90099d799da6051ade9feb5ac3e571f7d8cf2b8870e
Message ID: <01BB160E.8178A020@gate.bcdev.com>
Reply To: N/A
UTC Datetime: 1996-03-20 10:48:41 UTC
Raw Date: Wed, 20 Mar 1996 18:48:41 +0800

Raw message

From: Blake Coverett <blake@bcdev.com>
Date: Wed, 20 Mar 1996 18:48:41 +0800
To: "'Alan Olsen'" <alano@teleport.com>
Subject: RE: Microsoft's "answer" to Java
Message-ID: <01BB160E.8178A020@gate.bcdev.com>
MIME-Version: 1.0
Content-Type: text/plain


-----BEGIN PGP SIGNED MESSAGE-----

Commenting on a quote from a magazine about MS's new 
code-download/wintrust stuff Alan Olsen wrote:

>As a web developer, I have some problems with this scheme.  Giving Microsoft
>access to virtually every OLE control on the Web does not make me more
>secure.  Sounds like a way to rip off ideas from the rest of the development
>world.  If someone has a control that might compete with a Microsoft
>product, it could be shelved and/or delayed for "further security testing".

I think you've been badly misled on this one.  I've just been through all of the 
related specs from the MS INetSDK. While they are still incomplete in places, 
they look pretty workable to me.  In particular the certainly don't suggest that
MS would be involved in signing anything.  To quote from the beta docs:

        The present tools therefore allow any user of this development release to
        authorize themselves as a "Software Publisher" for test purposes and to
        sign their code, allowing for extensive testing of the tools and code
        used but not actually providing a secure infrastructure. In future releases,
        the tools will require software publishers to obtain certificates from
        companies whose function is to verify the identity of the publishers,
        providing end-users with a high level of assurance about the authenticity
        and origin of code that they receive.

>Java has a decentralized mechanism for security.  No one group controls what
>is a "certified" control and what is not.  You write the code and compile it
>and that is that. Furthermore, you are not stuck with Microsoft approved
>platforms.  (I wonder if there will ever be a version of Explorer for the Mac.)

The current version (2.0) is already available on the Mac and the 3.0 alpha 
versions appear to be about equally buggy on both the Win32 and Mac
platforms.  (I haven't, on the other hand, heard any news of Unix versions.
Perhaps Bristol and/or Mainsoft will cover that port.)

================

Here's my quick overview of the specifications in question for those interested.

Microsoft is providing the following components:

- - A generic trust management called (ever originally) WinTrust.
  WinTrust provides an API to ask whether a given subject is trusted to perform
  a specific action.  The API is extensible in that multiple 'Trust Providers'
  can be installed and each can define the types of subjects and actions they
  manage.  The docs define the role of a 'Trust Administrator' who can configure
  the rules used by the trust provider services be neglect the give the details.
  
- - An implementation of a trust provider called the 'Windows Software Publishing
  Trust Provider'
  This provider supports subjects which are executable images and the action
  of 'being published software'.  The decision to trust is based on a PKCS #7
  embedded within the executable containing a signed digest from the author
  and a chain of X.507 certs back to some configurable set of CAs.  If the
  executable is not verified the user is prompted with the offer to approve it manually.

- - A set of developer tools for creating your certificate and signing executables.
  Note that the beta includes a hard-coded root CA key and all certs must
  trace back to it.  The existing library for munging executable images has
  also been enhanced to support adding, removing, enumerating and retrieving
  certs from an image as well as reading the stream that should be included in
  digest calculations.

- - A single function solution for browsers and other applications to download,
  verify, install, and create a class factory for an OLE object given an URL.
  In the web case the HTML <OBJECT> tag is used to embed an OLE object 
  in a page.  The browser tries to create it based on the CLSID attribute (which
  contains a DCE-ish uuid.)  If it fails it calls CoGetClassObjectFromURL() passing
  in the URL from the CODE attribute of the same tag.  This function does all the
  magic including the WinTrust call from above.  (Apparently there will also be
  support for an 'Internet Search Path' if the CODE attribute isn't specified.)

Then of course there is the MS CryptoAPI but that's a discussion for another day.

- -Blake (who hasn't worked for Microsoft for years now)
-----BEGIN PGP SIGNATURE-----
Version: 2.6.3i
Charset: noconv

iQCVAwUBMU/Dirmr67p11D8rAQGHnQP/YI+EjCIcpBF3HQznruVBUkGsZls1ZVTf
SRvPJN7n+HrtvQ4WFSyAawsPnhRH183GTrtWAy+yhmmuzA6/Br/+rNJ/q0jSIlZw
w+RUsni9H9a7NsO1Y9xPQq//SHODYC0K+1vB6tU8XE56lZf9F0IZ4iP4El4PUWxD
7kXMboN1Nf0=
=5eH2
-----END PGP SIGNATURE-----






Thread