home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!nntp1.radiomail.net!csl.sri.com!dstine
- From: dstine@cisco.com (David Stine)
- Newsgroups: comp.protocols.appletalk
- Subject: Re: Extended- vs. nonextended-net net range aging
- Date: 23 Nov 1992 08:58:38 GMT
- Organization: cisco Systems, Menlo Park, California, USA
- Lines: 82
- Distribution: comp
- Message-ID: <1eq6fuINNar6@roche.csl.sri.com>
- References: <KAI.92Nov22234334@pyrite.SOM.CWRU.Edu>
- NNTP-Posting-Host: regal.cisco.com
-
- In article <KAI.92Nov22234334@pyrite.SOM.CWRU.Edu> kai@pyrite.SOM.CWRU.edu (Kai Getrost) writes:
- >[...]
- >
- >The problem:
- >
- > Inside AppleTalk (2nd ed.) gives slightly different algorithms for
- >expiring THIS-NET on extended vs. nonextended nets.
- >
- > On nonextended nets, it says (p. 5-18/19) that the value of
- >A-ROUTER and THIS-NET are gleaned from every RTMP Data packet
- >received. If no RTMP packets are received for 50 seconds, A-ROUTER
- >should be expired (set to 0), but *not* THIS-NET.
- > On extended nets, however, it says that *both* A-ROUTER and THIS-NET
- >should be reset when the aging timer expires.
- >
- > In other words, nodes on nonextended nets will save their link's net
- >number if all routers go down, but nodes on extended nets are supposed
- >to reset theirs (and set them again once RTMP packets are received
- >again, possibly to a new value). Why do these algorithms differ?
-
- Remember that on extended cables, you can have a _range_ of network numbers.
- The router could come back on the net with a different network number in its
- address.
-
- Example: let's say that you have an extended AppleTalk Ethernet with a
- cable-range of, oh, 10-20.[footnote 1] Let's say that the router's current
- net.node address is 10.240. Now, let's say that there is a hiccup in the power
- and when the router comes back on-line, all non-reserved node numbers in
- network 10 have been taken. The router, while AARP'ing for an open address,
- must choose a node in network number 11. Say 11.20.
-
- If it were the _only_ router on the cable, the nodes which were holding an
- A-ROUTER value and not timing it out would still be sitting on the wire, deaf
- to the world, since they are waiting for the router to re-appear on network
- 10.
-
- The analgous scenario occurring on a non-extended cable means that the nodes
- will have to re-AARP for new node addresses, since the cable configuration has
- changed. On an extended net, the nodes' current address is still valid and the
- cable configuration hasn't changed; simply the address of the router has
- changed.
-
- > The benefit of expiring the net range on extended nets could be
- >that you can reconfigure the net range by taking down just the routers
- >for some minimum time, and then bring them up with a new value that
- >all nodes will (hopefully) adopt. But this will affect any current
- >connections non-router nodes have, as they change to the new net range
- >and adopt new net:node IDs, and Inside AppleTalk is nebulous as to
- >when and how these connections should be reset (since it's a rare
- >occurence and difficult to deal with effectively).
-
- Yes, this is one of many instances where "Inside AppleTalk" is rather, well,
- vague. Perhaps wishy-washy. A tad economical with the facts, even.
-
- In addition to reading Inside AppleTalk on this issue, you should also be
- reading the document from APDA called:
-
- "AppleTalk Phase 2 Protocol Specification, C0144LL/A"
-
- which will set you back $20. I don't have my copy here right now (I'm at home)
- but this document contains the details on router disappearance and
- re-appearance that are missing from "Inside AppleTalk." If you are
- implementing an AppleTalk node or router, you must have this document. Simple
- as that.
-
- No, I'm not connected or getting a kick-back from APDA in any way. ;-)
-
- dsa
-
- [Footnote 1]
- There is a heap of misconception out there as to cable-range configurations on
- Phase II routers. I frequently see people with a cable-range 10 to 100 wide
- when they have an Ethernet segment with only 100 nodes _max_.
-
- Apple forgot to stress the point that a "wide" cable range (ie, a cable range
- which would be 10-11, rather than 10-10) is required _only_ when you will have
- more than 253 nodes on that particular cable/media segment. This usually only
- occurs in large bridged LAN's. If there are note going to be more than 253
- AppleTalk nodes on said cable segment, there isn't any point in configuring a
- wide cable range.
-
-
-