Re: [pim] What is the value of 'infinity'

["Christopher Thomas Brown" <ctbrown at nexthop.com>]

John Zwiebel initiated a hyperspace jump with:
> 
> On Apr 18, 2006, at 1:59 PM, Christopher Thomas Brown wrote:
> 
> >
> >In PIM-SM the infinity value for metric preference is 31-bit so its
> >0x7fffffff.
> 
> Only in the old spec which is replaced by the new which doesn't say.

There is only 31-bits in the preference field in the assert message.
The 32cd bit is the RPT bit. So preference is a 31-bit value.
Look at the preference field in the assert message in the PIM-SM
specification.

> And since the old spec doesn't give a clue as to why 31 bits vs 32
> or even vs only 8, I'm looking for a concensus to make the pronouncement
> "official".

This is conjecture on my part. Some routing protocols can have large
metrics. ISIS, for example, has extended metrics that can be up to 29 or
so bits. I don't know of the timing between when that was added to ISIS
and the PIM assert message was designed. Perhaps it was forward looking
or it was known.

So, with the metric field being 32-bits it makes sense to add another
32-bits to the message size to preserve 32-bit alignment. One bit was
used for the RPT bit. That leaves 31-bits for preference. As to why 31
over 16 or 8, perhaps because some routing implementations have
preferences that are that large.

> FWIW: I believe that in most cases its isn't going to matter.
> And in the bidir interop problem I found, there are some fairly
> obvious work-arounds.  But in the end, it would be better if
> we all agreed.
> 
> When the new PIM-SM replaces 2362, it is kind of harsh to expect
> new implementations to comply with an obsolete spec.

I think PIM-SM is mostly clear although I think it would have been
better if the spec spelled out the actual value of inifinity for both
metric and preference instead of relying on implication from the size of
the fields in the assert message.

I'm not a fan of a spec just waving its hands and say look here.
Especially for something that is just a paragraph or two of text.  IMHO,
the bidir spec should just spell out how to do the comparison.

Chris

_______________________________________________
pim mailing list
pim at ietf.org
https://www1.ietf.org/mailman/listinfo/pim