[ubuntu-uk] Remote Repos

[Alan Pope]

Hi Ian,

Terse reply, I am on 9600 baud on holiday - ssh, don't tell the wife!

On Fri, Jul 27, 2007 at 10:35:07PM +0100, Ian Pascoe wrote:
> Next, why would anyone want it?  I think the main areas would be:
> 
> * Unable to get access  regularly to a high speed Internet connection
> * unable to get access outside of an Intranet
> * no Internet connection at all
> 

Agreed.

> Additionally, there is also the small business, or school, that don't want
> their bandwidth gobbled up by machines on the network each individually
> updating themselves.

Yes, a location with multiple machines that all need updating/installing can 
get a head start with a "repo in a box". They can then keep that up to date.

> with a chip set that we know functions well with Ubuntu / Linux.  From this
> you'll realise I'm looking at each HDD having one architect for one release
> cycle on it.  My only question is whether a USB v2.0 or v1.1 or mixture;
> probably v2 only for availability reasons and data transfer speeds.
> 

It's rare to find mass storage devices that Linux cant use these days. Get a 
USB 2.0 one and it will be backwards compatible with USB 1.1 and USB 1.0. 

> Next is the physical side of things.  As Alan says you need a central server
> with about 400 - 500 Gb on it with the ability to handle updating at least 2
> HDDs at a time. 

It can update multiple disks at a time, you can plug multiple disks into the 
various USB ports (or a switch). I dont think keeping the disks up to date 
is an issue myself. It's just a case of running apt-mirror regularly and 
then rsyncing the master down to each USB disk. 

> I don't think more is needed as the uptake in the UK would
> be fairly minimal but at least the service is there.  The ideal locations
> would be in developing countries.  So the question here is whether to look
> at sending items from the UK, getting local offices set up through the local
> or nearby Loco teams, or just limit ourselves to the UK and Eire.
>

You'd be surprised how many people in the UK have ropey internet 
connections. Especially those who cannot afford the upgrade to Vista, or 
high speed broadband. But yes, I agree there are plenty of non-UK places who 
could benefit from this.
 
> Next is the physical location of the business.  Well that actually doesn't
> matter that much; in fact it could be quite easily run out of the proverbial
> spare bedroom.
> 

Somehwere with good net connectivity is all really.

> This of course also leads onto the question of how you actually ship them
> out to the client and get them back again.  I realise that the HDDs are
> designed to be portable but you still would probably need to have some
> specialised packing to protect them whilst in transit that could be reused
> to send it back to the "spare bedroom".
> 

The box they come in would be a good start, but specialised containers which 
survive multiple postings would possibly be better. Look at how other people 
ship stuff for return. 

> Lastly, is how do we let people know that the service exists?  Ideally, it'd
> be from the official Ubuntu sites, and also from those sites linked to
> Ubuntu and Linux.
> 

Marketing the service is another subject, yes. I can think of plenty of ways 
to pimp a service like this.

> So that's my thoughts on the actual project, but what about the competition?

Well, anyone can do this, anyone with a fast connection and a little money 
to initially put into the project. I envisaged doing this myself as a 
private enterprise, but there's nothing stopping anyone else doing the same.

> Other questions which spring to mind are:
> 
> * client gets their HDD, how do they pursuade apte-get or whichever to use
> that repo?

A sample sources.list and some documentation would be on the hard disk.

> * finances

I have thrown together spreadsheet with details of costings and it 
certainly isn't a "fast buck" kind of business.

> * licencing

This is something I worried about a bit. The potential inclusion of the 
multiverse (and other 3rd party repos such as medibuntu) would make the 
whole thing more usable and attractive but may be a legal minefield. 

> * sole tradeing or LLP for the business

Or if you're already an existing LTD you're sorted - there are a few LTDs on 
this list (myself included).

> * binaries / source code or both

This could be a selectable option. User just ticks a box or specifies in an 
email that they would like the source and this doubles (roughly) the size of 
the repo they get.

I'll be thinking about this a bit more this week whilst on holiday :)

Cheers,
Al.