I was discussing an IS-IS question with Giuseppe on the WAN forum, and it raised a sharper question in my mind about the true meaning of "adjacency" in both IS-IS and OSPF networks. As the books tell us, on a broadcast network; OSPF routers form adjacency only with the DR and BDR. This can be taken to mean that adjacency means "FULL" state. DRothers will have the DR and BDR in their ospf neighbor tables as "FULL" and fellow DRothers as "2WAY."
The books tell us that IS-IS routers on a LAN form adjacency with all other routers on the LAN; but still only the DIS is responsible for sending LSP's. If that's so, what is the meaning of two non-DIS routers on the LAN being "adjacent" with each other; if they're not swapping routing info?
Or are they in fact swapping routing info (LSP's) with each other, and the DIS is only responsible for sending out the LSP for the pseudonode? Similar to a summary LSA in OSPF?
The term "adjacency" seems to mean different things, and it's tripping me up on the BSCI practice quizzes...
Okay, I see that the "gory details" about IS-IS are necessary... sorry for reading your original post too hastily.
If your book says that only the DIS is responsible for sending LSP's, it is partially correct. The behavior can be described as follows:
1.) If a non-DIS router needs to transmit a LSP, it just does it directly by multicasting it onto a segment. There will be no explicit acknowledgement from any of the neighboring routers. Only the DIS will include the new LSP in its CSNPs. If the router sees its LSP in the subsequent CSNPs periodically sent by the DIS, it will know its LSP has been transmitted successfully to the DIS and hopefully to the remaining routers on the segment. The CSNPs sent by the DIS serve an implicit acknowledgement purpose in this case.
2.) When a router needs to request a LSP (perhaps it is a new router that has just come onto a multiaccess segment, or just happened to somehow miss the LSP sent by other router), it request it by multicasting a PSNP. This PSNP will also be received by all routers on a segment, however, only the DIS will respond with the requested LSP. This way, unnecessary flooding of that LSP is prevented on a segment. This is probably what your book talks about.
In short, if an update is sent, it is sent to everybody on a segment, not just to DIS. And in the opposite direction, if an update is requested, it will be answered by the DIS. I hope this makes it a little more clear now.
The meaning of adjacency is similar in OSPF and IS-IS, however, their usage and the following consequences differ significantly.
Two routers are said to be adjacent if they are allowed to exchange routing information directly. For a multiaccess segment, in OSPF, every new information (that is, the LSU with a number of LSAs) has first to be sent to the DR/BDR, and only afterwards, the DR will resend this information back to the segment so other routers know about it. By this procedure, OSPF makes sure that the DR/BDR are first to know about any change in the network topology.
The IS-IS is different. The DIS has a much weaker role in LSP flooding and link-state database synchronization. If a router on a multiaccess segment wants to announce a topology change, it simply sends a multicasted LSP onto the segment and every router will receive and process it. There is no notion of "DIS must be the first to know about everything" in IS-IS. There are some special responsibilities of IS-IS when helping newcoming routers synchronize their databases or when acknowledging received LSPs but they aren't so important right now.
So you got it right! The adjacencies are utilized differently in OSPF and IS-IS. The DIS in IS-IS is primarily responsible for creating a pseudonode LSP but it also has a supporting role in database synchronization on a multiaccess segment (don't want to go into gory details right now) but his operation is much simpler than that of an OSPF DR.
Note that the pseudonode in IS-IS is in fact almost the same as the LSA2 in OSPF. They both serve the same purpose - they simplify the description on a multiaccess network so that the link-state database is simpler. Regarding the pseudonode, you might want to have a look on this thread:
http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&forum=Network Infrastructure&topic=WAN%2C Routing and Switching&topicID=.ee71a06&fromOutline=&CommCmd=MB%3Fcmd%3Ddisplay_location%26location%3D.2cd43a87
While the topic of that thread is slightly different, it tries to explain the idea of LSA2 and pseudonode and the difference it makes in a network.