subdomains for testing things

[Daniel Pocock]

On 08/02/18 11:02, Max Mehl wrote:
> # Daniel Pocock [2018-02-08 11:00 +0100]:
>> There is a distinction between people volunteering to maintain a service
>> and the association choosing to rely on a service.
>>
>> This is particularly important in cases where two services do something
>> similar (e.g. Discourse acts as an alternative to the existing Mailman
>> service).  If half the group uses one service and half the group uses
>> the other, you split the organization or you double the amount of effort
>> required to community.  Metcalfe's law[1] comes to mind.
>
> I have to disagree in this case, with the positive experiences from the
> Git service [^1] in mind. Neither Discourse nor Gitea are/were
> officially planned to act as a replacement for any service.
> Git was something a few community members have wished for, and Björn and
> me just set it up. We were happy that it didn't entail any huge

Git is designed from the ground up as a distributed tool so that is
vastly different.  I would love to see a mailing list alternative that
uses a distributed architecture like Git as a back-end (although it
would still be up to the group to decide on using it)

Each project that uses Git can do so without impacting other projects.

Communication tools (Mailman, Discourse, XMPP) are a special case though
because everybody needs to use them.


> bureaucrazy [sic], we were able to make some tests right away, and to
> invite some people to give us feedback. That way we experienced that
> Gitea can also act as a replacement for SVN in the future and fits
> nicely in some workflows of our organisation. To make it official, we
> just had to announce it, no domain change, no votes of huge groups.
>
> Discourse could work similarly. It has been set up by a group of
> volunteers and we gave them a free hand. Later it might serve as a
> communication platform for a specific campaign or activity, and if we
> will make good experiences, other groups and parts of the organisation
> might think about picking it up for their activities, potentially now as
> an "official" service. There is no need that we *now* think about
> replacing the GA mailing list.
>


But it is not that simple.  If you start using it for a campaign, you
are either
a) forcing everybody who interacts the campaign to use it too, or
b) isolating the campaign from the rest of the community.

Neither is ideal.

Consider the impact by Metcalfe's Law, imagine we have 200 volunteers
using a single communication tool for all campaigns:

Value = 200^2 = 40,000

Now imagine if you have 150 volunteers using email and 50 using Discourse:

Value = 150^2 + 50^2 = 25,000

What Metcalfe's Law is telling us is that an organization committed to a
single platform is stronger than an organization that spreads itself
over different platforms.  It works either way: even if 150 volunteers
switch to Discourse and only 50 remain on email, the organization is
still weaker.

Note that in this example, I'm ignoring all the other differences
between the platforms (e.g. forums like Discourse are more prone to
censorship and tampering) and only focusing on the strength of the network.

It is also worth remembering that FSFE needs to communicate with people
beyond the community: once again the global email network has a value
with Metcalfe's Law, but each forum instance is like a little island.



> In my experience, bureaucracy frustrates volunteers for very good
> reasons. Let them define a subdomain name, let them hack around, give
> them some freedom – if such a service ever is ready for
> organisation-wide usage, we can still think about the details. But
> devaluate their service by putting a "test" in the domain name would
> demotivate me as a service maintainer and user at the same time.
>

In fact, we have a similar thing in Debian: any Debian Developer can set
up subdomains under debian.net almost instantly but only officially
supported things go under debian.org.  This strategy has been very
successful and is well understood in the Debian community.

Something can only migrate from debian.net to debian.org if there is
widespread consensus about it.

It is about the community consciously deciding which direction to go and
also being willing to support something even if the volunteers who
started it drop out.


> And, once again, your proposal solves a non-problem in my opinion.
> hellekin found harsh words to express his feelings, but his problem
> wasn't that the service implied to be official but that he didn't know
> about it.
>

Well, I hope my calculation with Metcalfe's Law helps people understand
why it is a problem, at least when we talk about communication tools.

> Best,
> Max
>
>
> [^1]: git.fsfe.org
>