The IP Over NBMA (ION) Archive

["Timothy J. Smith" <tjsmith@VNET.IBM.COM>]

>> Date: Wed, 18 Dec 1996 11:07:21 -0800
>> From: rajeev@trillium.com
>> To: ion@nexen.com, kvm@trillium.com, bgleeson@cisco.com
>> Subject: Re:  1577 client vs classic2 server: interoperablity issues
>>  [...]
>> If not, then we basically have a non backward compatible spec and
>> system administrators will have to explicitly configure their
>> LISs to run either all v1 or all v2. Is that the intended "transition
>> from v1 to v2?

>What is the liklihood of vendors providing only minimal implementations
>of v1, (i.e., that the vendors never enhanced their v1 server
>implementation)?

>What is the liklihood that a customer installed an early, minimal v1
>server implementation and then never upgraded?

I think the issue here is not whether the vendors and administrators
will try to do a good job.  I believe the problem will be one more of
timing.  Vendors obviously are not all on the same schedule, and in
a multi-vendor site, you may very well have a situation where some
of the releases of code are V1 and some are V2.  In addition, in
large sites, it becomes almost impossible to reload everyone's
software at once.

>By the way, I suspect that vendors will do the "right thing" for their
>customers without cluttering up the classic2 and NHRP specs.  I think
>that, for example, mandatory InaATMRPs, ought to _not_ be included in
>classic2.

I agree with keeping the spec as clean as possible, but I think
an additional sentence might solve this problem.  How about this:

"Given that older clients may rely on the server sending them inATMARP
requests as the registration mechanism, an ARP Server MAY send out
inATMARP requests over a VCC that the client has established."

Cheers,
TJS (the other one)