[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [Debconf-discuss] no debconf wireless in Davis


sorry for the late mail in this thread, but this wasn't said here before, and 
IMO is still worth saying. (As we had discussions about video teams cabling 
needs basically every year...)

On Dienstag, 3. August 2010, Ben Hutchings wrote:
> On Tue, 2010-08-03 at 09:12 +0200, Andreas Tille wrote:
> > On Tue, Aug 03, 2010 at 01:06:16AM -0400, Holger Levsen wrote:
> > > the debconf ESSID was introduced in Davis 8h before Debian Day started.
> > > Unfortunatly it breaks the video, both streaming and recording.
> > >
> > :-( I know.  However, I can confirm that Marks talk yesterday at
> >
> > 22:00-23:00^W^W 16:00-17:00 in Davis was perfectly fine while the talk
> > after this by Jorge O. Castro was terrible.
> >
> > > There is time to improve this situation on wednesday, when no talks are
> > > scheduled.
> >
> > So no better situation today? :-(
> > Is there any reason to not use a cable for the video streaming?????
> There is a cable.  The GuruPlug was used to shape traffic to ensure that
> most of its bandwidth was reserved for video streaming.  However, there
> appears to be some problem with the configuration that causes the TCP
> connections to be reset.

Also there were several problems with "this cable":

- it (=the guruplug) was set up 8 hours before the talks started on Debian 
Day, so we (=videoteam) had no time to test+fix+change things. 

- then there was a DOS attack happening in the .10 network of the CS 
department, which in turn turned off random ports on their switches to debug 
things. our port was among those...

- then there was an IPv6 configuration problem with the guruplugs, leading to 
multicast traffic from the big auditorium messing up the network in the small 
talk room upstairs (!!)

- some things I probably forgot (ignorance is bliss!)

While in the end we managed to deal with all those problems, to me the 
conclusion of these events is clear: the videoteam needs its own ethernet 
cables for the whole video network. And, this isnt a new conclusion, we 
reached it already in 2006, before DebConf6. And now as we had the failures 
we wanted to avoid, we have a proof that this conclusion was right. Thanks 
for that ;-p

We need "our own cables" because despite we actually use TCP/IP over ethernet, 
(for large parts) we dont use it like this, but rather as a transport medium 
for firewire, which is not nearly as fault tolerant as tcp/ip.

So thats why we need our own _cables_. A managed vlan is a bad compromise, 
esp. if the management is (partly or fully) done outside debconf (like the CS 
department). If the management is done by us, videoteam or debconf-admins at 
the location, it can work. (Like it did for dc9.)


Attachment: signature.asc
Description: This is a digitally signed message part.

Reply to: