#104 Behavior if no AKD response is received

[Wesley Tanaka]

Date: Sun, 30 Sep 2001 12:01:37 -0400
From: "Tessa Lau" <tlau@ofb.net>
Subject: Re: gale akd requests

Dan, did you reply to this?

My take is that a key could have been "revoked" and
thus depending on
the cache could result in incorrect behavior.

On the other hand it could be a temporary network outage...

--Tessa

On Wednesday, Wesley Tanaka mumbled:
> in http://fugu.gale.org/protocol
>
> (2) claims that "If no positive replies are received,
the location does not
> exist and an error should be reported."
>
> gsend (contrary to this specification) appears to be
happy fetching keys from
> its cache, even if there would have been no response
to an AKD request.
>
> I'm reworking this part of praya and I was wondering
if it would be okay to n
>ot
> report any error if the key can be found in the
cache, without bothering to
> send out an AKD request.
>
> Thanks!
>
> Wesley

Date: Sun, 30 Sep 2001 09:03:53 -0700
From: "Dan Egnor" <egnor@ofb.net>
Subject: Re: gale akd requests

In most cases gsend will use the cache and not even
invoke AKD at all.
After a while, the cache times out, and it will start
invoking AKD (while
still using the cache to avoid delay). In that case,
if AKD fails, the
cache entry should be removed. I'm not sure if it is.

In other words, think of what's described in the
document as the "raw" AKD
protocol. The gale implementation happens to put a
cache in front of it.
As usual, using a cache means that you might be a
little bit out of date,
but that's the price you pay for performance (in the
absence of cache
callbacks).

Dan