On 7/30/2010 11:06 AM, Andrew Deason wrote:
> On Thu, 29 Jul 2010 10:21:41 -0400
> Jeffrey Altman <[email protected]> wrote:
> 
>> It is roughly correct.  The cell name is not transmitted as part of the
>> RPC to the pt server but when the foreign cell group entry is
>> constructed the name ends up exceeding the limit.
> 
> The ptserver doesn't allow user names to be longer than PR_MAXNAMELEN
> minus the length of the local cell name, which I think is the check
> that's failing... (CorrectUserName). This additional restriction is so
> the user can be used as a foreign user name in another cell, though;
> shouldn't we bypass that check for foreign users? We're not ever going
> to need to create a
> [email protected]@informatik.uni-goettingen.de, right?

This check is wrong for two reasons:

(1) The case you identified

(2) In modern usage of AFS, the foreign cell name is actually a
non-local realm name which does not need to match the cell name.

I think the test should be:

  if (localname)  {

     for each (localrealm)  {

        verify lengthof(n...@localrealm) <= PR_MAXNAMELEN

     }

  } else {

     verify lengthof(foreign-name) <= PR_MAXNAMELEN

  }

Jeffrey Altman




Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to