home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ferkel.ucsb.edu!taco!gatech!usenet.ins.cwru.edu!agate!doc.ic.ac.uk!uknet!gdt!ccsis
- From: ccsis@sunlab1.bath.ac.uk (Icarus Sparry)
- Newsgroups: comp.unix.admin
- Subject: Re: Netmask Questions
- Message-ID: <C1D6yD.6rr@sunlab1.bath.ac.uk>
- Date: 24 Jan 93 15:27:10 GMT
- References: <FPB.93Jan17104229@ittc.ittc.wec.com> <1jf55qINN9ai@early-bird.think.com>
- Organization: Bath University Computing Services, UK
- Lines: 67
-
- (Bogus distribution of 'comp' deleted).
- In the referenced article, barmar@think.com (Barry Margolin) writes:
- >In article <FPB.93Jan17104229@ittc.ittc.wec.com> fpb@ittc.wec.com (Frank P. Bresz) writes:
- >> The problem comes in the fact
- >> that the 3/280 lives on both the 129.228.8.XXX loop and the
- >> 129.228.112.XXX loop and I can't seem to set different netmasks for
- >> ie0 and ie2. i.e. it couldn't telnet to 129.228.8.238 or any high
- >> numbered 129.228.8.XXX machine.
- >
- >Most TCP/IP implementations, including Sun's, require all the subnets of a
- >network to use the same netmask. The RIP routing protocol, for instance,
- >doesn't include netmasks with the routes.
- >
- >> Is what I am trying just impossible? Is it unreasonable. The
- >> reasoning for this is that I can get 2nd ethernet cards for a
- >> SPARCStation cheaper than I can buy a bridge. Or I wouldn't bother
- >> routing packets I would just bridge the segments.
- >
- >You can use a Sun as a router. The problem is that you're trying to have
- >multiple levels of subnetting, and current implementations don't support
- >that. All subnets of 129.228 have to use the same subnet mask.
-
- Much as I hate to disagree with Barry, he is wrong. The proof is the network
- that we run here (138.38.x.x). He is correct in saying that you have
- problems with RIP, but we use static routing here for everything.
-
- Problems which you need to know about
-
- 1) Sun do not document how they select which interface they will use if
- more than one interface appears to be suitable. It turns out that they
- try the interfaces in order, and use the first one. This means that you
- must have the most restrictive netmask on the first (ie0 in this case)
- interface.
-
- 2) bootparamd is pretty stupid. It chooses an interface at random. I have
- a re-written version somewhere which does 'best match' rather than 'first
- match', which works.
-
- 3) You need to use static routing, but this is often a good idea as
- it places less load on the machines. In our case we have disk-less
- machines which are booted off the gateways, so if the gateway is
- down so is the machine, so the fact that it could get to our backbone
- is of little interest.
-
-
- =============================================== netmask 255.255.0.0
- | le2 138.38.32.33 | le2 138.38.32.34
- |++++++| |++++++|
- | sl1a |------ ----| sl1b |
- |++++++| le1 | |le1 |++++++|
- 138.38.35.2 | | | le0 138.38.35.65
- | le0 | | | netmask 255.255.255.192
- | =================|
- | | |
- |================= |
- | |
- netmask 255.255.255.192 |
- to slc-01 to slc-17 |
- to slc-18 to slc-35
-
- The 'le1' interface of sl1a has address 138.38.35.66, and the 'le1'
- interface of sl1b has address 138.38.35.1.
-
- So the solution to Franks original problem could be to turn off dynamic
- routing, and swap over the ethernet cables to his two interfaces.
-
- Icarus
-