-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi everyone,
I'm planning to shut down Tor Weather.
For those of you who have not used Tor Weather in the past, here's what it does: Tor Weather is an email notification service that lets users subscribe for notifications related to one or more relays. These notifications can be configured to be sent out when the relay unexpectedly goes down or when it pushes enough traffic to earn its operator a Tor t-shirt.
Unfortunately, Tor Weather is broken:
- It is not actively maintained, nor is there an active operator running the main instance on weather.torproject.org. If it breaks, we don't have a good plan for fixing it. But since it's not broken yet, nobody is actively working on developing a better Weather.
- The t-shirt notifications are misleading. Weather is promising t-shirts for relays that have not been around long enough or that provide too little bandwidth to be useful to the network, and it's almost impossible to deny a t-shirt after Weather has promised it. Also, Weather is likely not offering t-shirts to people who have long earned it, thereby confusing them.
- An unreliable notification system is worse than not having a system at all. Relay operators shouldn't rely on Weather to notify them when their relay fails. They should rather set up their own system instead.
There are a few possible alternatives to Weather:
- We started rewriting Weather using Onionoo as data back-end in 2014 [0]. That project didn't produce working code, but somebody could pick up these efforts and finish the rewrite.
- The Roster developers said that they're planning to include an email notification function in Roster [1].
- We developed a simple Python script that provides information about a relay operator's eligibility for acquiring a t-shirt [2].
None of these alternatives is a full replacement of Weather, and I don't think that we should work towards deploying an alternative anytime soon. The best scenario that I can imagine here is that, once Weather is shut down and people realize they really want to have something like it back, the relay operator community gets together and develops a thing or two. Those alternatives wouldn't run on torproject.org, but that seems secondary.
I'm thinking about the following timing for shutting down Weather:
- By end of month, March 31, we disable Weather's email notification functionality and place a big warning on the homepage that it will be shut down 2 months later. We also blog about this like we did when we shut down Tor Cloud [3].
- By June 30, we take the Weather host offline.
Please respond to this list by March 23, which is in 1 week from now, if you strongly object to this plan.
All the best, Karsten
[0] https://trac.torproject.org/projects/tor/wiki/doc/weather-in-2014
[1] http://www.tor-roster.org/
[2] https://gitweb.torproject.org/metrics-tasks.git/tree/task-9889/tshirt.py
[3] https://blog.torproject.org/blog/tor-cloud-service-ending-many-ways-remain-h...
On Wed, Mar 16, 2016 at 06:51:10PM +0100, Karsten Loesing wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi everyone,
I'm planning to shut down Tor Weather.
[...]
Unfortunately, Tor Weather is broken:
[...]
There are a few possible alternatives to Weather:
- We started rewriting Weather using Onionoo as data back-end in 2014
[0]. That project didn't produce working code, but somebody could pick up these efforts and finish the rewrite.
- The Roster developers said that they're planning to include an
email notification function in Roster [1].
- We developed a simple Python script that provides information about
a relay operator's eligibility for acquiring a t-shirt [2].
None of these alternatives is a full replacement of Weather, and I don't think that we should work towards deploying an alternative anytime soon. The best scenario that I can imagine here is that, once Weather is shut down and people realize they really want to have something like it back, the relay operator community gets together and develops a thing or two. Those alternatives wouldn't run on torproject.org, but that seems secondary.
Hi Karsten,
I know you said we should respond if we strongly object to your plan. I admit I don't object but I'm responding anyway.
Tor Weather is a great project, and it is one of the only ways The Tor Project actually communicates with relay operators and gives them feedback about their contribution. I'm very much looking forward to Roster acting as a replacement, especially if it is integrated with the t-shirt system. But, with that being said, I am saddened that we're getting rid of the service without having ready a drop-in replacement. However, considering the current problems, it seems wise that the service is shutdown before the situation becomes worse.
I think I prefer a service like this run by the Tor Project, but considering all relay contact information and its uptime/bandwidth/etc are all public, a third-party service wouldn't receive much additional personal information, anyway.
I'm thinking about the following timing for shutting down Weather:
- By end of month, March 31, we disable Weather's email notification
functionality and place a big warning on the homepage that it will be shut down 2 months later. We also blog about this like we did when we shut down Tor Cloud [3].
The one problem I foresee with this date is that it is nearly 1 April and that date is associated with jokes and pranks in some parts of the world. So, if this association is something we care about then maybe Weather should be disabled a few days earlier or later.
- By June 30, we take the Weather host offline.
That day seems like its as good a day as any.
Thanks for running Weather this long. Hopefully someone(s) will produce a working replacement soon.
Thanks, - Matt
[0] https://trac.torproject.org/projects/tor/wiki/doc/weather-in-2014
[1] http://www.tor-roster.org/
[2] https://gitweb.torproject.org/metrics-tasks.git/tree/task-9889/tshirt.py
[3] https://blog.torproject.org/blog/tor-cloud-service-ending-many-ways-remain-h...
Hello Karsten,
Do you think rewriting/maintaining/fixing Tor Weather would make a nice GSoC project? What I mean by this is that do you think it can be three months of work at least, with scope for future improvements to be carried out later?
If yes, we should propose this as a GSoC project and then see if we can find someone to mentor it.
We at Roster are willing to commit to a satisfactory Roster replacement. A GSoC for doing so would be welcome.
-V
On Thursday, 17 March 2016, Sukhbir Singh azadi@riseup.net wrote:
Hello Karsten,
Do you think rewriting/maintaining/fixing Tor Weather would make a nice GSoC project? What I mean by this is that do you think it can be three months of work at least, with scope for future improvements to be carried out later?
If yes, we should propose this as a GSoC project and then see if we can find someone to mentor it.
-- Sukhbir _______________________________________________ tor-project mailing list tor-project@lists.torproject.org javascript:; https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-project
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 17/03/16 06:39, Virgil Griffith wrote:
We at Roster are willing to commit to a satisfactory Roster replacement. A GSoC for doing so would be welcome.
Great, I'm looking forward to trying out this Roster feature once it's available.
Of course, that shouldn't stop people from writing their own Weather-like thing.
All the best, Karsten
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 17/03/16 05:37, Sukhbir Singh wrote:
Hello Karsten,
Do you think rewriting/maintaining/fixing Tor Weather would make a nice GSoC project? What I mean by this is that do you think it can be three months of work at least, with scope for future improvements to be carried out later?
If yes, we should propose this as a GSoC project and then see if we can find someone to mentor it.
Fine questions. Maybe take a look at the Weather in 2014 wiki page:
https://trac.torproject.org/projects/tor/wiki/doc/weather-in-2014
I'm not enough of a web front-end developer to say how hard it is to rewrite Weather. But I think the back-end can be easily solved by using Onionoo's data instead of writing a Tor controller and keeping an own relay database.
I think three months is sufficient to produce something useful.
The question of finding a mentor might be more difficult to answer. I would offer to be backup mentor and also to answer questions about Onionoo. But we'd still have to find a primary mentor. How about you? :)
All the best, Karsten
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 17/03/16 05:05, Matthew Finkel wrote:
On Wed, Mar 16, 2016 at 06:51:10PM +0100, Karsten Loesing wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi everyone,
I'm planning to shut down Tor Weather.
[...]
Unfortunately, Tor Weather is broken:
[...]
There are a few possible alternatives to Weather:
- We started rewriting Weather using Onionoo as data back-end in
2014 [0]. That project didn't produce working code, but somebody could pick up these efforts and finish the rewrite.
- The Roster developers said that they're planning to include an
email notification function in Roster [1].
- We developed a simple Python script that provides information
about a relay operator's eligibility for acquiring a t-shirt [2].
None of these alternatives is a full replacement of Weather, and I don't think that we should work towards deploying an alternative anytime soon. The best scenario that I can imagine here is that, once Weather is shut down and people realize they really want to have something like it back, the relay operator community gets together and develops a thing or two. Those alternatives wouldn't run on torproject.org, but that seems secondary.
Hi Karsten,
Hi Matt,
I know you said we should respond if we strongly object to your plan. I admit I don't object but I'm responding anyway.
Tor Weather is a great project, and it is one of the only ways The Tor Project actually communicates with relay operators and gives them feedback about their contribution.
Well, to be fair, the main purpose of Onionoo and all its clients is to give relay operators feedback about their contribution. The Onionoo page contains an (incomplete) list of clients:
https://onionoo.torproject.org/
And Weather could be listed there as well if somebody changes it to use Onionoo data instead of keeping its own database. It's not that we're abandoning all our efforts there just by shutting down Weather.
I'm very much looking forward to Roster acting as a replacement, especially if it is integrated with the t-shirt system. But, with that being said, I am saddened that we're getting rid of the service without having ready a drop-in replacement. However, considering the current problems, it seems wise that the service is shutdown before the situation becomes worse.
Agreed. I would have preferred having a replacement before shutting it down, but I wanted that for over two years now and the replacement did not materialize.
I think I prefer a service like this run by the Tor Project, but considering all relay contact information and its uptime/bandwidth/etc are all public, a third-party service wouldn't receive much additional personal information, anyway.
Right, it's all public data.
I'm thinking about the following timing for shutting down Weather:
- By end of month, March 31, we disable Weather's email
notification functionality and place a big warning on the homepage that it will be shut down 2 months later. We also blog about this like we did when we shut down Tor Cloud [3].
The one problem I foresee with this date is that it is nearly 1 April and that date is associated with jokes and pranks in some parts of the world. So, if this association is something we care about then maybe Weather should be disabled a few days earlier or later.
Huh, you're right. Let's move that date to April 4 then.
- By June 30, we take the Weather host offline.
That day seems like its as good a day as any.
Right. I'll also ask our sysadmins if they prefer an earlier date.
Thanks for running Weather this long. Hopefully someone(s) will produce a working replacement soon.
Yep!
Thanks, - Matt
Thanks for your response!
All the best, Karsten
[0] https://trac.torproject.org/projects/tor/wiki/doc/weather-in-2014
[1] http://www.tor-roster.org/
[2] https://gitweb.torproject.org/metrics-tasks.git/tree/task-9889/tshirt.py
[3]
https://blog.torproject.org/blog/tor-cloud-service-ending-many-ways-remain-h...
_______________________________________________
tor-project mailing list tor-project@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-project
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hello list,
On 17/03/16 09:22, Karsten Loesing wrote:
On 17/03/16 05:05, Matthew Finkel wrote:
On Wed, Mar 16, 2016 at 06:51:10PM +0100, Karsten Loesing wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi everyone,
I'm planning to shut down Tor Weather.
I'm moving forward with shutting down Tor Weather. I'm planning to make three changes today:
1. Disable new subscriptions by removing the "Sign up for Tor Weather!" link from the Tor Weather homepage.
2. Replace the home page text with the text on the following pad and pastebin (it's trivial to change this text, so if people here have feedback on making the text better, just let me know, and I'll incorporate that any time):
https://pad.riseup.net/p/jDF3DxkIj7EU
http://paste.debian.net/423803/
3. Disable welcome messages for newly joining relays.
I'll also send a message to tor-relays@ with the same text about discontinuing Tor Weather.
After June 30, 2016, I'll remove all dynamic parts from Tor Weather, including the email sending parts, and replace the website with a single static HTML page.
All the best, Karsten
tor-project@lists.torproject.org
-
Karsten Loesing -
Matthew Finkel -
Sukhbir Singh -
Virgil Griffith