Re: About address allocating (IPv6, variable length SLA/prefixes?)
Hi Francis et.al., as I said a couple of days ago in Budapest, I would like to see an explanation and/or review from the routing point of view. Judging from my (limited) knowledge about IPv6, going for a variable length SLA field would either leave us with "wasted" address space (as the network next door would be a different site and thus should have a different NLA field anyway), or we would end up with a variable length network prefix length (much like in the v4 environment), effectively extending the NLA field into the SLA field. Doing so would probably require a cross-check against existing IPv6-aware IGPs. That is where I would like to see input from the routing camp(s). Regards, Wilfried. ______________________________________________________________________ From: Francis Dupont <Francis.Dupont@enst-bretagne.fr> To: itojun@iijlab.net CC: Haisang Wu <hswu@ns.6test.edu.cn>, 6bone@ISI.EDU Subject: Re: About address allocating Date: Sat, 20 May 2000 15:17:57 +0200 In your previous mail you wrote:
hi, I have the following questions about address allocating: I know SLA is /48, and interface ID should be 64 bits, does it mean that the smallest unit when allocating address is /48? In other words, if I allocate a /48 to a large university, could I allocate a /48 to four middle schools, thus each middle school gets an block less than /48, which is /50. Is this plan reasonable?
=> we'd like to get a /48, ISPs would like to give a /64 to us: - /48 seems a bit too large for a default allocation size - /64 is unusable when you need subneting then the current idea, as presented yesterday here in Budapest at the RIPE meeting, is to introduce "small site" which get /56 (on byte boundary, large enough for up to 256 subnetworks or a few levels of hierarchy). Then /56 will become the default allocation size in RIR allocation & assignment document.
I'm not sure if introducing "small sites" is a good thing... when we switch ISP and they force me to switch from /48 to /56, renumber becomes very hard. => the idea is that it is easier for someone which needs a /48 to deal with its ISP than for a common customer to fight in order to get a /48 because /64 is not enough: this is a compromise for common customers (ie you at home, IIJlab is strong enough to get a /x with x <= 48). I believe it is a good compromise... Regards Francis.Dupont@enst-bretagne.fr -------------------------------------------------------------------------------- _________________________________:_____________________________________ Wilfried Woeber : e-mail: Woeber@CC.UniVie.ac.at UniVie Computer Center - ACOnet : Tel: +43 1 4277 - 140 33 Universitaetsstrasse 7 : Fax: +43 1 4277 - 9 140 A-1010 Vienna, Austria, Europe : RIPE-DB: WW144, PGP keyID 0xF0ACB369 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Hi, On Sat, May 20, 2000 at 06:59:23PM +0200, Wilfried Woeber, UniVie/ACOnet wrote:
Doing so would probably require a cross-check against existing IPv6-aware IGPs. That is where I would like to see input from the routing camp(s).
The routing protocols *have* to be able to work with a arbitrary-length prefix anyway (in the /0 to /64 range). After all, you might want to use the same routing protocol in the NLA "area" and inside a customer network (SLA space). Gert Doering -- NetMaster -- SpaceNet GmbH Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Wilfried, Architecturally, IPv6 has the equivalent of variable length subnet masks built in. There are really only two boundaries that are not flexible- the boundary between the format prefix and the rest of the address, and the /64 boundary. (The format prefix is in fact variable length, but it is architecturally defined.) So any IGP or EGP design needs to be fully flexible to the left of /64. Subnetting to the right of /64 would be tricky. Brian "Wilfried Woeber, UniVie/ACOnet" wrote:
Hi Francis et.al.,
as I said a couple of days ago in Budapest, I would like to see an explanation and/or review from the routing point of view.
Judging from my (limited) knowledge about IPv6, going for a variable length SLA field would either leave us with "wasted" address space (as the network next door would be a different site and thus should have a different NLA field anyway), or we would end up with a variable length network prefix length (much like in the v4 environment), effectively extending the NLA field into the SLA field.
Doing so would probably require a cross-check against existing IPv6-aware IGPs. That is where I would like to see input from the routing camp(s).
Regards, Wilfried. ______________________________________________________________________ From: Francis Dupont <Francis.Dupont@enst-bretagne.fr> To: itojun@iijlab.net CC: Haisang Wu <hswu@ns.6test.edu.cn>, 6bone@ISI.EDU Subject: Re: About address allocating Date: Sat, 20 May 2000 15:17:57 +0200
In your previous mail you wrote:
hi, I have the following questions about address allocating: I know SLA is /48, and interface ID should be 64 bits, does it mean that the smallest unit when allocating address is /48? In other words, if I allocate a /48 to a large university, could I allocate a /48 to four middle schools, thus each middle school gets an block less than /48, which is /50. Is this plan reasonable?
=> we'd like to get a /48, ISPs would like to give a /64 to us: - /48 seems a bit too large for a default allocation size - /64 is unusable when you need subneting then the current idea, as presented yesterday here in Budapest at the RIPE meeting, is to introduce "small site" which get /56 (on byte boundary, large enough for up to 256 subnetworks or a few levels of hierarchy). Then /56 will become the default allocation size in RIR allocation & assignment document.
I'm not sure if introducing "small sites" is a good thing... when we switch ISP and they force me to switch from /48 to /56, renumber becomes very hard.
=> the idea is that it is easier for someone which needs a /48 to deal with its ISP than for a common customer to fight in order to get a /48 because /64 is not enough: this is a compromise for common customers (ie you at home, IIJlab is strong enough to get a /x with x <= 48). I believe it is a good compromise...
Regards
Francis.Dupont@enst-bretagne.fr -------------------------------------------------------------------------------- _________________________________:_____________________________________ Wilfried Woeber : e-mail: Woeber@CC.UniVie.ac.at UniVie Computer Center - ACOnet : Tel: +43 1 4277 - 140 33 Universitaetsstrasse 7 : Fax: +43 1 4277 - 9 140 A-1010 Vienna, Austria, Europe : RIPE-DB: WW144, PGP keyID 0xF0ACB369 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Brian E Carpenter Program Director, Internet Standards & Technology, IBM On assignment for IBM at http://www.iCAIR.org Attend INET 2000: http://www.isoc.org/inet2000 Non-IBM email: brian@icair.org
participants (3)
-
Brian E Carpenter -
Gert Doering -
Wilfried Woeber, UniVie/ACOnet