To:
"Hollenbeck, Scott" <shollenbeck@verisign.com>
CC:
ietf-provreg@cafax.se
From:
Daniel Manley <dmanley@tucows.com>
Date:
Wed, 19 Sep 2001 12:40:08 -0400
Sender:
owner-ietf-provreg@cafax.se
User-Agent:
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.4) Gecko/20010913
Subject:
Re: Command Recovery
yuck, enforcing unique client trid's since the begining of operations could get relatively costly for each transform transaction, even if uniqueness is only required within a registrar. How about enforcing uniqueness within a day? The transaction info/check command could require a clTRID and a date to find the object. Dan Hollenbeck, Scott wrote: >>-----Original Message----- >>From: Daniel Manley [mailto:dmanley@tucows.com] >>Sent: Wednesday, September 19, 2001 11:31 AM >>To: ietf-provreg@cafax.se >>Subject: Re: Command Recovery >> > > >>I guess option 1 seems like the most versatile and useful, so my >>preference is with 1. Could client trids still be optional if you went >>ahead with number 1? Does it even make sense to keep them optional in >>that case? >> > >I think they can still be optional, you just lose the ability to check up on >one if it hasn't been provided. > >I think a bigger question involves uniqueness: does the server have to start >enforcing unique TRIDs for each client? Across all clients? As things are >written right now, the clTRIDs are basically just passed-through data that >the client can use to synchronize commands and responses, and the server >logs them as part of the larger transaction ID. If the server doesn't >enforce uniqueness you can get multiple hits when trying to determine >status, which may not be helpful. If the server does enforce uniqueness, it >becomes a bit more complex. > ><Scott/> >