IETF
sidr
sidr@jabber.ietf.org
Thursday, 28 July 2011< ^ >
melkins has set the subject to: SIDR WG http://www.ietf.org/proceedings/79/agenda/sidr.html
Room Configuration

GMT+0
[12:48:55] Wes George joins the room
[12:56:05] EuiJong Hwang joins the room
[12:58:00] Carlos Martinez joins the room
[12:58:12] pscanlon joins the room
[13:00:47] Warren Kumari joins the room
[13:02:35] <Carlos Martinez> Good morning everyone
[13:03:22] Benno Overeinder joins the room
[13:03:44] David Cooper joins the room
[13:04:10] Ricardo Patara joins the room
[13:04:42] <Carlos Martinez> I will be your jabber scribe today
[13:04:47] Shane Amante joins the room
[13:05:02] Bert Wijnen joins the room
[13:05:07] Geoff Huston joins the room
[13:05:39] weiyinxing joins the room
[13:08:51] <Carlos Martinez> Sandra (WG Chair) currently going thorough the documents status'
[13:10:02] Arturo Servin Ü joins the room
[13:10:05] <Carlos Martinez> Roque Gagliano with draft algorithm agility
[13:10:12] <Carlos Martinez> update on the −02 version
[13:11:09] jpc joins the room
[13:11:49] <Carlos Martinez> ready for WGLC ?
[13:11:52] <Carlos Martinez> questions? comments?
[13:12:38] <Carlos Martinez> Now on to Rob Austein
[13:12:47] <Carlos Martinez> RPKI Publicaction Protocol
[13:12:57] <Carlos Martinez> -01 version
[13:14:08] <Carlos Martinez> different security constraints for certs/ROAs
different availability constraints for certs/ROAs
[13:15:36] <Carlos Martinez> publish without running a 24/7 system but holding to your private keys
[13:15:49] <Carlos Martinez> [protocol overview]
[13:16:27] <Carlos Martinez> recent protocol changes slide
[13:16:52] <Carlos Martinez> now operations are generic: <publish> <withdraw>
[13:17:02] <Carlos Martinez> tracking other documents
[13:17:23] <Carlos Martinez> questions? comments?
[13:17:40] weiler joins the room
[13:17:50] <Carlos Martinez> q from wg chair: isn't this the same as the provisioning protocol ?
[13:18:06] <Carlos Martinez> now on to Terry Manderson
[13:18:26] <Carlos Martinez> draft ietf sidr usecases -02
[13:18:51] <Carlos Martinez> updates slide
[13:19:12] <Carlos Martinez> no feedback received
[13:19:22] lochii joins the room
[13:19:33] <Carlos Martinez> authors ask for WGLC
[13:19:55] <Carlos Martinez> [presentation ends]
[13:20:06] <Carlos Martinez> questions or comments ?
[13:20:31] <Carlos Martinez> mic: 2 calls for WGLC ?
[13:20:39] <Carlos Martinez> mic: Rob clarifies he did not ask for WGLC
[13:21:00] <Carlos Martinez> Now on to Matt Lepinski
[13:21:05] <Carlos Martinez> draft bgpsec protocol
[13:21:33] jpc leaves the room
[13:21:46] <Carlos Martinez> what is bgpsec ? slide 2
[13:22:14] <Carlos Martinez> slide 3, about the draft's status
[13:22:44] <Carlos Martinez> slide 4, related documents
[13:24:07] <Carlos Martinez> slide 5, two new documents
[13:24:47] <Carlos Martinez> slide 6, mailing list discussion
[13:26:27] <Carlos Martinez> presentation ends
[13:26:58] <Carlos Martinez> mic: on algorithms, how does this interact with the work done on algorithms for rpki ?
[13:27:31] <Carlos Martinez> presenter: it is not clear at this point if the same algorithms are the best choice for bgpsec
[13:28:38] <Carlos Martinez> Now on to Randy Bush
[13:29:05] <Carlos Martinez> BGPsec beaconing for replay reduction
[13:29:21] <Carlos Martinez> slide 2: replay attack
[13:30:27] <Carlos Martinez> slide 3: why reply attack ?
[13:31:17] <Carlos Martinez> slide 4: replay reduction
[13:32:04] <Carlos Martinez> origin re-announces prefix well within that lifetime, hence *beaconing*
[13:32:28] <Carlos Martinez> slide 5: origination by as0 to as1
[13:32:51] <Carlos Martinez> slide 6: announcement as1 to as2
[13:33:42] <Carlos Martinez> slide 7: replay elimination is a non-goal, goal is to *reduce* the vuln window
[13:34:01] <Carlos Martinez> slide 8: protocol, not intent
[13:35:31] <Carlos Martinez> slide 9: why multi beacon
[13:36:50] <Carlos Martinez> slide 10: believe only previous TTL
[13:38:51] <Carlos Martinez> slide 11: so believe minimum TTL
[13:39:56] <Carlos Martinez> slide 12: alternatively...
[13:41:11] <Carlos Martinez> slide 13: what it costs
[13:42:00] <Carlos Martinez> slide 14: bottom line
[13:42:52] <Carlos Martinez> [presentation ends]
[13:43:21] <Carlos Martinez> mic: are there potential security concerns if someone decides to drop the beacons repeatedly long enough for them to expire ?
[13:43:37] <Carlos Martinez> mic: just to be sure we are covering that
[13:43:52] <Carlos Martinez> presenter: it is suggested that times are set appropiately
[13:44:04] <Carlos Martinez> presenter: thus accidental drops are covered
[13:44:16] <Carlos Martinez> presenter: this does not cover monkey in the middle attacks
[13:45:46] <Carlos Martinez> comments questions from the jabber room?
[13:46:19] <Carlos Martinez> Now on to Bert Wijnen
[13:46:27] <Carlos Martinez> RPKI MIB Modules
[13:46:44] <Carlos Martinez> two mib modules
[13:46:51] <Carlos Martinez> rpki-origin-validation-mib
[13:46:53] <Carlos Martinez> rpki-router-mib
[13:47:00] <Carlos Martinez> (slide 2)
[13:47:37] <Carlos Martinez> slide 3; bgpValROATable
[13:49:05] <Carlos Martinez> slide 4: RPKI router peer table 1/2
[13:50:00] <Carlos Martinez> slide 6: next steps
[13:50:44] <Carlos Martinez> - do we want to add new stuff to bgp4-MIB
[13:50:45] <Carlos Martinez> ?
[13:51:46] <Carlos Martinez> slide 7: next steps 2/2
[13:52:35] <Carlos Martinez> authors ask for WG adoption
[13:53:13] <Carlos Martinez> mic: in previous protocols when the protocols are new, mibs come along, and mibs are not adopted as there is no experience
[13:53:31] <Carlos Martinez> mic: project was started by one router vendor
[13:54:20] <Carlos Martinez> mic: the roa table is populated by an external cache, this is used mainly to show uses data, do you really want people altering your validation status via SNMP ?
[13:54:45] <Carlos Martinez> mic: concern is whether to have a write version of the MIB
[13:55:38] <Carlos Martinez> mic: rob, we only want a read-only mib at this point on the server side, it should not be modifying anything
[13:56:04] <Carlos Martinez> mic: most operators believe that snmp is a simple monitoring protocol, not a management protocol
[13:56:59] <Carlos Martinez> question from the wg chair: ROATable… the description looks like the information is encapsulated in a ROA, and ROAs shoudl be validated, but the validity status is in the ROA Table
[13:57:21] <Carlos Martinez> mic: Rob: this is just a table with the cache data after validation
[13:58:06] <Carlos Martinez> wg chair: you might also want to capture whether the routes are valid or invalid
[13:58:29] <Carlos Martinez> mic: that should go into other mibs, do we also want to add this to other mibs ?
[13:58:52] <Carlos Martinez> mic: sandy you are right, that line does not belong there
[13:59:35] <Carlos Martinez> mic [rob]: there is no point in having that state there
[14:01:04] <Carlos Martinez> mic: randy: I support read-only. I think this particular question on the last line, i don't mind if it goes away
[14:01:42] <Carlos Martinez> ----
[14:01:48] <Carlos Martinez> Now on to Chris Hall
[14:01:54] <Carlos Martinez> Route Servers and BGPSec
[14:02:45] <Carlos Martinez> slide 2: route servers and internet exchanges
[14:02:45] Geoff Huston leaves the room
[14:03:52] <Carlos Martinez> slide 3: route servers and BGPSec
[14:04:34] <Carlos Martinez> request route server should be a requirement for the protocol
[14:04:42] <Carlos Martinez> slide 4: general requirements
[14:04:55] Roque Gagliano joins the room
[14:05:06] Geoff Huston joins the room
[14:05:17] <Carlos Martinez> slide 5: route server as proxy
[14:06:24] <Carlos Martinez> slide 6: route server signs for itself
[14:08:25] <Carlos Martinez> slide 7: otherwise?
[14:09:57] <Carlos Martinez> slide 8: in conclusion...
[14:11:09] <Carlos Martinez> mic: Randy: what is invisibility? i think what we really want is that the fact we use a route server should not influence the downstream bgp decision — it should not affect the overall bgp as path length calculation
[14:12:04] <Carlos Martinez> i don't think we can have invisibility and bgpsec
[14:12:10] <Carlos Martinez> transparency is fine
[14:12:31] <Carlos Martinez> the approach where the sending router shares it's private key to the RS is not possible
[14:12:54] <Carlos Martinez> mic: why does the RS have to be involved at all in the signature chain?
[14:13:19] <Carlos Martinez> presenter: it does not, but the signatures have to be there
[14:14:32] <Carlos Martinez> mic: the basis for the protocol is the signing
[14:14:54] <Carlos Martinez> mic: all the receiving as cares about is the validity of the signature
[14:15:18] <Carlos Martinez> wg chair: are you proposed that the shared signature should be taken out of the protocol , or just in this particular case ?
[14:15:51] <Carlos Martinez> are you suggesting a special case of the protocol ?
[14:16:26] <Carlos Martinez> mic: what does the destination as should be in that case /
[14:16:52] <Carlos Martinez> mic: i think the point about visibility is not inserting as in the path
[14:17:21] <Carlos Martinez> mic: invisibility should not be a concern
[14:17:34] <Carlos Martinez> mic: i dont't entirely follow option 3
[14:18:11] <Carlos Martinez> if you are willing to accept more changes to the bgp protocol, why not make the RS rewrites the data /
[14:18:55] <Carlos Martinez> mic: two obs, choice one we are giving one key and not THE key to the entire as
[14:19:04] <weiler> I think warren's suggestion was to have something new: rather than the RS pushing BGP, it pushes config info so rtrs can dynamically configuring peerings themselves.
[14:20:41] <Carlos Martinez> mic: randy the forward signing is absolutely neccesary
[14:21:49] <Carlos Martinez> mic: keyur patel, warren's addition is to make some sort of auto-discovery, so you replace a star with a full mesh
[14:22:30] <Carlos Martinez> wg chair: we should answer the question wher e RS support is a requirement
[14:23:25] <Carlos Martinez> rob austein: add some notifications could be a good addition
[14:24:08] <Carlos Martinez> mic, roque gagliano: one of the options was to have the RS forward-sign with its own AS
[14:24:55] <Carlos Martinez> you can flag a RS in a bgpsec certificate (draft cert profile)
[14:26:08] <Carlos Martinez> mic: who would attest to that ? (randy)
[14:26:27] <Carlos Martinez> randy: there are other ways to signal a RS
[14:26:42] <Carlos Martinez> randy: support for transparent RS should be a requirement
[14:27:31] <Carlos Martinez> randy: one of the reasons many people use RS is why they do not believe their routers scale to that many bgp sessions, so the signaling, autoconf approach runs into this problem
[14:27:33] <Carlos Martinez> ---
[14:28:19] <Carlos Martinez> Now on to Doug Montgomery, ptential optimizations for as-path prepending and transparent route servers
[14:28:30] <Carlos Martinez> slide 2: bgpsec-00 level set
[14:30:05] <Carlos Martinez> slide 3: bgpsec-00 Path_sigs
[14:31:06] <Carlos Martinez> slide 4: Optimizations / Enhancements
[14:32:33] <Carlos Martinez> slide 5: Prepending strawman
[14:33:35] <Carlos Martinez> requriement: protect prepends and not incur in the cost of multiple prepend signature
[14:34:13] <Carlos Martinez> slide 6: Transaprent route servers
[14:34:36] <Carlos Martinez> slide 7: Translucent RS strawman
[14:35:06] sigpipe joins the room
[14:36:22] <Carlos Martinez> apply the same idea of the prepend count with pCNT = 0 for RSs
[14:36:59] <Carlos Martinez> slide 8: further details
[14:37:54] <Carlos Martinez> [presentation ends]
[14:38:49] <Carlos Martinez> mic: some routers also support ingress as path prepending
[14:39:39] <Carlos Martinez> mic [ericsson]: two different as paths, one for data, other for messages
[14:41:18] <Carlos Martinez> mic [john scudder]: this attribute seems to be becoming the authoritative as_path. how does this impact loop supression?
[14:42:56] <Carlos Martinez> mic: if everybody in the as supports this new signature as the path, but as deployment would be incremental, other bgp speakers won't know about this attribute so as_path should be retained
[14:43:16] <Carlos Martinez> mic: what we do want in the signature, even if it adds redundancy, is to add the as#
[14:45:15] <Carlos Martinez> mic: i don't think protecting pretending is worth it
[14:46:13] <Carlos Martinez> mic [randy]: inbound as prepending is a lie
[14:46:31] <Carlos Martinez> to bias traffic for myself
[14:47:10] <Carlos Martinez> mic, john scudder: prepending is a form of policy
[14:47:32] <Carlos Martinez> there are other ways to influence policy that we have kept out of scope
[14:48:19] <Carlos Martinez> mic, danny mc pherson: if we are protecting prepending, then the counts should be protected, and the count validated
[14:48:36] <Carlos Martinez> i don't think we can do things that break policy
[14:48:40] <Carlos Martinez> other comments?
[14:48:59] <Carlos Martinez> ---
[14:49:39] <Carlos Martinez> Now on to Matthias Weisch
[14:49:53] <Carlos Martinez> Waehlisch (sorry)
[14:50:10] <Carlos Martinez> RTRLib : client-side implementation of RPKI in C
[14:50:26] <Carlos Martinez> slide 2: background, still working on beta version
[14:50:38] <Carlos Martinez> slide 3: architectural design
[14:51:21] <Carlos Martinez> slide 4: transport socket
[14:52:27] <Carlos Martinez> slide 5: RTR socket
[14:53:34] <Carlos Martinez> slide 6: PFX Table
[14:53:50] <Carlos Martinez> slide 7: RTR connection manager
[14:54:47] <Carlos Martinez> slide 8: Example - Establish connection
[14:55:01] <Carlos Martinez> slide 9: create server pool
[14:55:24] <Carlos Martinez> slide 10: create conn manager and perform origin validation
[14:56:16] <Carlos Martinez> slide 11: conclusion & outlook
[14:56:53] <Carlos Martinez> http://rpki.realmv6.org/
[14:57:30] <Carlos Martinez> question? comments?
[14:57:53] <Carlos Martinez> --
[14:57:57] adrianfarrel joins the room
[14:58:02] <Carlos Martinez> Now on to Sriram Kotikalapudi
[14:58:31] <Carlos Martinez> RIB Size Estimation for BGPSec
[14:59:09] <Carlos Martinez> slide 2: measurement of prefixes and paths in isps route reflectors
[14:59:11] <Carlos Martinez> and pe routers
[14:59:57] <Carlos Martinez> slide 3: update format and signature overheads
[15:00:38] <Carlos Martinez> slide 4: estimation of bgpsec update sites
[15:01:37] <Carlos Martinez> slide 5: bgpsec adoption rate
[15:02:13] <Carlos Martinez> slide 6: adoption rate (2)
[15:02:29] <Carlos Martinez> slide 7: pe router rib size estimation for bgpsec
[15:03:20] <Carlos Martinez> slide 8: route reflector rib size estimation for bgpsec
[15:03:42] <Carlos Martinez> slide 10: comparison rsa-2048 vs ecdsa-256
[15:04:44] <Carlos Martinez> slide 10: bgpsec sign algorithm transition example
[15:06:34] adrianfarrel leaves the room
[15:06:46] <Carlos Martinez> slide 11: bgpsec sign alg transition followed by proto transition
[15:07:03] <Carlos Martinez> slide 12: summary and conclusions
[15:07:51] adrianfarrel joins the room
[15:07:57] <Carlos Martinez> questions?
[15:08:18] <Carlos Martinez> mic: did you factor reducing the rsa keys smaller ?
[15:08:44] <Carlos Martinez> mic: does going to rsa-1024 reduce the 33 gig req by a large factor or not ?
[15:09:03] <Carlos Martinez> mic, randy: this is cheap ram
[15:09:21] <Carlos Martinez> the real problem is that routers are 32 bit archs and not 64 bit
[15:10:20] <Carlos Martinez> mic: a comments on ram being cheap: ram is cheap but space on control board is not
[15:11:21] <Carlos Martinez> Now on to Sriams's bgp-design-choices
[15:11:29] <Carlos Martinez> slide 3: comments / questions on the sider list
[15:11:50] mikeb joins the room
[15:12:09] <Carlos Martinez> slide 4: what is next for this document
[15:13:06] <Carlos Martinez> [presentation ends]
[15:13:52] <Carlos Martinez> Now on to Randy Bush
[15:14:26] <Carlos Martinez> [no slides at this point]
[15:14:31] <Carlos Martinez> - two uses for rpki
[15:14:40] <Carlos Martinez> [slides now on display]
[15:14:48] <Carlos Martinez> Resource Transfer in the Global RPKI
[15:15:27] <Carlos Martinez> slide 2: the goal
[15:15:48] <Carlos Martinez> transfer between two registries
[15:15:56] <Carlos Martinez> slide 3: the players
[15:16:09] <Carlos Martinez> slide 4: related parties
[15:16:45] <Carlos Martinez> slide 6: the simple swing
[15:17:50] <Carlos Martinez> slide 7: the sequence
[15:18:09] <Carlos Martinez> slide 8: indirect seller
[15:19:23] <Carlos Martinez> slide 9: indirect buyer harder
[15:20:52] <Carlos Martinez> slide 10: RIRs have problems
[15:21:51] <Carlos Martinez> slide 11: RIR cross proxy
[15:22:59] <Carlos Martinez> slide 12: even with the proxy scheme
[15:23:18] <Carlos Martinez> slide 13: perhaps the RIRs would like to share their plan
[15:23:25] <Carlos Martinez> [presentation ends]
[15:23:52] <Carlos Martinez> wg chair: are there problems that the wg should be working on ?
[15:24:28] <Carlos Martinez> randy: there is a 2007 document that describes a possible signaling protocol
[15:25:08] <Carlos Martinez> mic: which is the swing point?
[15:25:23] <Carlos Martinez> randy: it is the lowest point which agrees to be the swing point
[15:25:47] <Carlos Martinez> --
[15:26:01] Shane Amante leaves the room
[15:26:06] <Carlos Martinez> Now on to draft zhang-idr-decoupling
[15:26:13] <Carlos Martinez> slide 2: comments
[15:26:47] EuiJong Hwang leaves the room
[15:27:40] Shane Amante joins the room
[15:28:18] <Carlos Martinez> Incremental deployment evaluation
[15:28:19] Shane Amante leaves the room
[15:28:35] <Carlos Martinez> [presentation ends]
[15:28:51] <Carlos Martinez> wg chair: can you summarize the changs in the new version of the draft ?
[15:29:01] <Carlos Martinez> presenter: some clarifications relating to the comments in the ML
[15:29:18] <Carlos Martinez> --
[15:29:37] <Carlos Martinez> Now on to Yang Xiang
[15:29:43] <Carlos Martinez> FS-BGP
[15:29:51] Warren Kumari leaves the room
[15:30:03] Geoff Huston leaves the room
[15:30:07] <Carlos Martinez> Fast and Secure BGP
[15:30:46] <Carlos Martinez> slide 3: FS-BGP
[15:30:58] Bert Wijnen leaves the room
[15:31:12] <Carlos Martinez> slide 4: terminology
[15:31:58] weiyinxing leaves the room
[15:31:59] <Carlos Martinez> slide 5: terminology (2)
[15:32:33] <Carlos Martinez> slide 6: signatures in s-bgp
[15:32:54] <Carlos Martinez> slide 7: pros and cons of s-bgp
[15:33:02] David Cooper leaves the room
[15:33:34] adrianfarrel leaves the room
[15:33:56] <Carlos Martinez> slide 9: announcement restrictions in bgp
[15:34:25] Benno Overeinder leaves the room
[15:34:32] Benno Overeinder joins the room
[15:35:14] <Carlos Martinez> slide 10: critical path segments
[15:35:31] Ricardo Patara leaves the room
[15:36:09] <Carlos Martinez> slide 11: fs-bgp: CSA
[15:36:38] <Carlos Martinez> slide 13: cost-reduction
[15:38:09] <Carlos Martinez> slide 15: CSA achieves feasible path authentication
[15:38:51] Roque Gagliano leaves the room
[15:38:52] <Carlos Martinez> slides 16: forge a path in fs-bgp is possible
[15:39:39] Roque Gagliano joins the room
[15:39:56] <Carlos Martinez> conditions of effective hijacking
[15:40:04] <Carlos Martinez> (that was slide17)
[15:40:25] <Carlos Martinez> slide 18: prevent effective hijacking
[15:40:51] <Carlos Martinez> slide 19: SPP suppressed path padding
[15:41:11] <Carlos Martinez> slide 21: security level
[15:41:29] <Carlos Martinez> slide 22: computational cost
[15:41:46] weiler leaves the room
[15:41:49] <Carlos Martinez> mic: are we proposing an alternative solution ?
[15:42:22] <Carlos Martinez> mic: can we re-spin it in terms of merging the work with the existing proposals as optimizations
[15:42:59] <Carlos Martinez> mic (s. kent): when we re-chartered the group we agreed that we rejected feasible path solutions
[15:43:11] <Wes George> sorry, previous commenter was me - forgot to announce
[15:43:16] <Carlos Martinez> as they don't reflect how bgp really works
[15:43:20] <Carlos Martinez> thanks wes
[15:43:59] Wes George leaves the room
[15:44:11] <Carlos Martinez> wg chair: can you get to a conclusion slide?
[15:44:18] <Carlos Martinez> presenter:
[15:44:21] Arturo Servin Ü leaves the room
[15:44:26] <Carlos Martinez> call for wg adoption
[15:44:49] <Carlos Martinez> [presentation ends]
[15:45:23] <Carlos Martinez> wg chair: are you proposing we re-consider the position of the wg ?
[15:45:36] <Carlos Martinez> additional charter items
[15:45:54] <Carlos Martinez> wg chair: i do not know if this could be an additional charter item
[15:46:22] <Carlos Martinez> a humm is called
[15:46:33] Benno Overeinder leaves the room
[15:46:34] <Carlos Martinez> result is that it won't be included in the carter of the wg
[15:46:38] Roque Gagliano leaves the room
[15:49:28] Carlos Martinez leaves the room
[15:52:34] sigpipe leaves the room
[16:03:09] Warren Kumari joins the room
[16:04:37] Warren Kumari leaves the room: Replaced by new connection
[16:04:58] Warren Kumari joins the room
[16:08:05] Shane Amante joins the room
[16:16:58] Warren Kumari leaves the room
[16:21:14] Shane Amante leaves the room
[16:59:38] Bert Wijnen joins the room
[17:01:07] adrianfarrel joins the room
[17:03:47] adrianfarrel leaves the room
[17:04:16] Warren Kumari joins the room
[17:23:10] Arturo Servin Ü joins the room
[17:23:30] Arturo Servin Ü leaves the room
[17:24:07] Arturo Servin Ü joins the room
[17:28:16] lochii leaves the room
[17:29:03] Arturo Servin Ü leaves the room
[17:31:46] Warren Kumari leaves the room
[17:42:00] Warren Kumari joins the room
[18:01:51] jpc joins the room
[18:03:05] mikeb leaves the room
[18:09:28] mikemlb joins the room
[18:09:43] mikemlb leaves the room
[18:31:50] jpc leaves the room
[18:32:00] Warren Kumari leaves the room
[18:47:07] Roque Gagliano joins the room
[18:48:58] Roque Gagliano leaves the room
[18:50:15] Roque Gagliano joins the room
[18:54:15] Bert Wijnen leaves the room
[19:07:46] Warren Kumari joins the room
[19:07:55] Warren Kumari leaves the room
[19:24:31] Bert Wijnen joins the room
[19:28:16] Warren Kumari joins the room
[19:29:44] Warren Kumari leaves the room
[19:39:46] Roque Gagliano leaves the room
[19:47:15] Warren Kumari joins the room
[20:05:12] Warren Kumari leaves the room
[20:39:56] pscanlon leaves the room
[22:20:23] Bert Wijnen leaves the room
[23:40:55] Bert Wijnen joins the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!