pem-dev
[Top] [All Lists]

Re: CRLs Load with PCAs and CAs

1993-11-09 15:17:00
Bob,

In your message to Anish you went from suggesting that CA-based 
CRL responders would be nice, to creating a requirement to be able 
to derive a user's mailbox address from his DN.  This is NOT a 
requirement for PEM; it is a problem that comes with the 
particular CRL-responder proposal that you and Sead and some 
others have put forth.  Any discussion of the need to be able to 
map DNs into mailbox addresses should take place only in the 
hypothetical context of that proposal.  The thought of skewing DNs 
used in PEM to fix this problem introduced by this proposal is 
really stretching things!

You ask about the accuracy required of timestamps for CRLs, and I 
believe that even a Mickey Mouse watch is sufficient, given the 
human time scale in which compromise reporting, etc, takes place.  
Again, as for having the CA provide time stamping services, one 
must question the trustworthiness of this in a larger context.  If 
an employee at company A sends an organizational message to 
someone at company B, I think company B would like an independent, 
third-party to provide time stamping of the message and, if 
needed, of the relevant CRLs.  Also, it is often important that a 
timestamp be applied at the request of the recipient, not the 
originator.  The rationale is that it is sometimes (often?) more 
important to record when a message was received and the recipient 
acted upon it, rather when the originator claims to have sent the 
message.


Steve

=========================== and another response ==================

To: jueneman%wotan(_at_)gte(_dot_)com
Subject: Re: Corporate Identity and Authorization
Cc: pem-dev(_at_)tis(_dot_)com
------------

Bob,

I want to comment on some of the points in your response to 
Charlie, message.  First, although Apple OCE does not initially 
have CRL support, I am confident it will be provided shortly.  So 
this seems to be a minor issue.

The PEM RFCs were revised two years ago (in I-Ds, not so recently, 
really) to remove the DN restrictions.  This change was intended 
to avoid unduly constraining DN selection by folks who were 
establishing directory services and who, presumably, knew enough 
about X.500 to construct "good" DNs.  I disagree with your premise 
that ANY syntactically valid DN is fair game.  The intent of the 
change was to not preclude semantically and syntactically valid 
DNs.  By the way, the list of valid DN attributes is not to be 
maintained by CCITT or NIST OIW but rather by the IANA.  I'll try 
to get that done soon with an eye toward avoiding some of the 
X.500 attribute types that are inappropriate for DNs, though 
appropriate for X.500 entries.  Also, the 1422 rule about display 
of unknown attribute applies to approved attributes that are just 
not implemented yet, not arbitrary attributes.

Bob, there are folks who manage X.500 DSAs and who understand what 
constitutes a reasonable DN.  Just because people (who do not have 
real X.500 experience) have had trouble sorting out syntactically 
valid constructs from semantically valid ones, this does not mean 
that the use DNs is intrinsically kludgey.  

I sympathize with your desire to let your secretary read your mail 
when you are away, but I question whether everyone wants to have 
ALL of their mail read by their secretary under such 
circumstances.  PEM was not designed to accommodate either 
"selective reading by alternative recipients" capability in an 
elegant fashion.  You propose one approach which relies on 
separation of signing and encryption key management material, but 
that does not solve the problem of the user who wants his 
secretary to read most of his mail, but not all of it.  That 
problem might be solved by using two different certificates, with 
the second having a DN in which the term "secretary" is included 
in a predictable fashion as a naming convention.  Under this 
convention, mail could be sent directed to the secretarial 
certificate unless it was especially sensitive.  This is an 
alternative solution which can be accomplished without changing 
the PEM specs and it solves both the problem you cited and the 
other one I noted.  Should PEM be changed to use dual certificates 
to accommodate your solution approach work?  Is my dual 
certificate hack better or worse than your perversion of DNs?  I 
know which way I judge these, but that's a subjective judgment!


Steve
=============== and yet another response ===================

To: jueneman%wotan(_at_)gte(_dot_)com
Subject: Re: DNs (Re: Corporate Identity and Authorization)
Cc: pem-dev(_at_)tis(_dot_)com
------------

Bob,

In your response to Ali I think you overloaded the utility of a 
directory.  The DN DOES describe the entity to whom the 
certificate is issued, NOT the certificate!  The directory model 
does not assume that we can completely trust the information 
stored therein.  The directory makes explicit provision for 
separate entries for organizational and residential persons, 
organizational roles, etc.


Steve


<Prev in Thread] Current Thread [Next in Thread>