Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!umich!samsung!brutus.cs.uiuc.edu!psuvax1!psuvm!PUCC!GETTES From: GETTES@PUCC.BITNET (Michael R. Gettes) Newsgroups: bit.listserv.nodmgt-l Subject: Re: Routing tables from pathalias Message-ID: Date: 9 Feb 90 15:35:42 GMT Sender: Node Management Discussion Reply-To: Node Management Discussion Organization: Princeton University, CIT Network Systems Lines: 60 Approved: NETNEWS@PSUVM Gateway In-Reply-To: Your message of Thu, 08 Feb 90 12:04:22 EST On Thu, 08 Feb 90 12:04:22 EST you said: >I guess I need to be more clear in the way that I say things. >Obviously I knew that node number information is available from other >sources. The question should have been stated as "Why is this >information needed to generate JES2 routing tables not included?" I >assume that the answer is probably that pathalias does not allow for >it. Couldn't it be bundled in with the name, though? As in something >like acadia-1234? As Mr. Hrybyk noted, it could easily be added to a generic table using his specifications in a config file. I also think it would be a trivial modification to include the nodenumber in the generic routing table of base pathalias. Now, before Ed gets annoyed that I am meeting Roger on this point for pathalias, the reason I feel the inclusion of nodenumbers is important in a generic table is that the symmetry of BITNET is brought about by the nodenumbers. In other words, this is how we break ties and can be important information during topological analysis, and would greatly simplify some of the analyzing tools I have developed for myself. So, from the standpoint of routing and analysis, I feel nodenumbers in a generic table is justified. Additionally, I think I can argue that if we did not need nodenumbers to break ties on BITNET, then the :NODENUM tag is no longer necessary since this too would be a "local" configuration problem. If I am wrong on this -- I am sure Roger or LDW will be quick to correct me ;-). >More generally, my question was intended to point out that generating >routing tables from the above described pathalias output is sometimes >more involved than a simple textual transformation. This may be true. And maybe not. I have already written something to do the RSCS tables in REXX. It was easy. Adding some additional information to the problem will not make the problem an order of magnitude greater. >A second question that I asked Michael offline is "Why would anyone >want to spend any time at all writing such a program when GENROUTS is >already available?" Well, I feel I have described quite a number of the merits of pathalias. I will include below what I had answered to Roger offline. >The current GENROUTS is no good. The new GR uses a yet to be proven >algorithm for generating routing tables. Pathalias has been in use >for years. It works great! It is already in C, only one version needs >to be maintained. If we remove the function of formatting a routing table >from GENROUTS, then we no longer have to worry about people getting the >right version of GENROUTS to generate a specific table. Every time a >new format is required, because somebody writes a new NJE emulator in >a fishbowl and dreams up some new file format, then GENROUTS will have >to be modified to handle it. Then everyone must get a copy of the new >program to be up to level since it is possible to slip in changes to >the routing algorithm. Everyone is affected by this change. This way, >if a new table format is needed, then just write a post-processor for >the pathalias map. You could then give that program to some central >clearing-house, like BITNIC, so others having your system would not >have to re-invent the wheel. >Lastly, the generic table is a map of routing, something >GENROUTS does not produce and will not produce in the new program. >These pathalias maps are extremely useful. /mrg