Re: Please give a summary (Was: Debian services and Debian infrastructure)
On 21/01/14 at 18:11 +0000, Ian Jackson wrote:
> Lucas seems to be intending to mediate these offers to interested DDs
> (who have Debian-related uses for a VM) directly, with the apparent
> expectation that those DDs would end up administering those VMs
> directly in an ad hoc manner. DSA haven't been involved or informed
> (until now).
That's not accurate, sorry.
I mentioned that I was planning to do that in
In the same mail, I asked DSA if they would be willing to manage those
> I can see at least three problems with this, which have been mentioned
> in this and previous threads:
> Firstly, there is the prospect that "bad things" would happen to these
> VMs. For example, they might get compromised; or access to them could
> be lost when the invididual DDs who had been running them leave or go
> on vacation. This would be bad for the project, and of course it's
> bad for DSA because in such a situation DSA would be asked about these
> VMs and expected to fix it but have no access to or knowledge about
> Secondly, there is the risk that there would be no coherent way to
> retire these VMs when they are no longer needed. When we take on
> ongoing donated resources like that, there should be a mechanism for
> ensuring that the project knows about them, can periodically check
> that they're still being used and needed, etc.
Sure, it's important that the project's resources, whatever they are,
are properly tracked and managed.
> Thirdly, it increases the risk of services being developed in a way
> that would make them hard to deploy on DSA-managed infrastructure.
> Developers of services would benefit from early contact with DSA to
> understand at least in general terms how to make a readily deployable
> and maintainable online service.
Interesting. How does it *increase* the risk of services being developed
in a way that would make them hard to deploy on DSA-managed
infrastructure? The current status is that people host their services on
their own hardware. I don't see how moving services to VMs would make
On the contrary, if done properly, this provides a chance to monitor
which services are being developed, review designs, etc.