Provide a forum for the MIMIC community about mimic-website HOT 34 CLOSED

One option for the forum is:
https://www.discourse.org/about/

For an example, see:
http://discuss.gohugo.io

from mimic-website.

I don't think we can roll out the discussion forum right away, since we'll have to be thoughtful about the user management system. I'm hesitant to add yet another signup/user system. It should either integrate with something existing (which Discourse seems to allow with google/github signup) or integrate with the PhysioNet user system (which would be a serious overhaul).

In the meantime, would the github issue tracker be sufficient for now?

from mimic-website.

We're getting snowed under with emails from people using the new version of MIMIC and we haven't announced it yet, so I'd like to get something in place fairly soon (definitely before we announce MIMIC formally).

The trouble with sticking with GitHub is that we already know that it's not an ideal solution. If we postpone the decision to move, we'll end up creating more work in migrating content and more disruption for the community.

I'm not committed to Discourse, but I do think we should make a decision pretty soon. Can we try to reach a decision by the end of the week? To get things moving, I'll set out my reasons for using Discourse below:

• Nothing will be permanent, so our best option is to choose something well-used with a reasonable chance of being around in 2 or 3 years.
• Discourse has strong pedigree. It was started by the founder of StackExchange. It also has some major users, like Twitter (https://twittercommunity.com), imgur (https://community.imgur.com), and GitHub (https://discuss.atom.io).
• It is open source, free, and should be relatively easy to escape from if and when we move to a different solution in the future.
• It integrates with user accounts for Github, Twitter, Google, Facebook, etc so user management is taken care of already.

Please could people add alternative suggestions here? I'm sure there are other good options. We can then (hopefully!) come to a group decision.

from mimic-website.

Hey @alistairewj @Franck-Dernoncourt @tnaumann @spitz-dan-l @ikarosilva ... does anyone have experience with user forums? We're looking for a place to support discussion within the MIMIC community. Trying to work out what our options are!

from mimic-website.

Unsure, I know many libraries use Google Groups for the mailing list, but the interface is quite ugly and inconvenient to use.

There is a list Stack Exchange clones here: http://meta.stackexchange.com/a/37953/178179

Among the forums I frequent:

• OpenStack uses Askbot: https://ask.openstack.org/en/questions/# but it's also hurting my eyes :)
• MetaOptimize uses OSQA: http://metaoptimize.com/qa/ I prefer it is visually over Askbot. Looks like metaoptimize is slow at the moment, even timeout sometimes, https://ask.wireshark.org/ is another OSQA example.

I don't have any experience with Discourse but I had also heard of it, that sounds like a reasonable option.

from mimic-website.

Good suggestions, thanks Franck. A StackExchange-style system would be a nice way to go, so it's a shame that there don't seem to be any polished, free ones. Of the free StackExchange clones on the list, Askbot looks like the best bet (http://askbot.org/en/questions/).

from mimic-website.

By the way, speaking of community knowledge and documentation, I found this proposal interesting: Warlords of Documentation: A Proposed Expansion of Stack Overflow.

from mimic-website.

No idea about which solution is "best". Obviously stack overflow is huge but - at least for me - most of the time I'm redirected to a stack overflow question/answer through google. So in that respect the actual site layout of the stack overflow website may or may not be ideal (I've never actually used it) - but their question/answer subpages are obviously superb and heavily utilised.

The biggest issue is the user registration. Honestly the problem is we have this PhysionetWorks registration which is tedious to manage and completely disconnected from the rest of the web. I don't think anyone on the team has the bandwidth to overhaul it though. In my opinion, our best bet is to stick with GitHub account integration with everything else. That way people can keep their work life distinct from their social media if they so please, but it's not necessary since anything that integrates a GitHub account undoubtedly integrates the more common social media accounts. In this light I'd go with discourse - has great pedigree as you say and seems to have the features we want: simple, can be organized into categories, has GitHub integration, open source, and it's something we can easily maintain.

from mimic-website.

Discourse looks nice, but I'm not sure it would be "easily" maintained. Anything we have to install ourselves runs the risk of becoming outdated and unmaintained (see joomla, redmine). In my brief scan, it seems to require ruby/rails? Are you guys comfortable with that? I suppose we'd be able to host in on AWS, but there is the ongoing maintenance.

Same probably goes with any of the other Open Source QA projects.
I've used Question2Answer ( http://www.question2answer.org/ ) previously, but it quickly fell victim to disuse and spam.
OSQA (does look cleaner) and Askbot seem similar, but I can't find anything post-2013 that actually compares them.

It would be nice if we could start a mimic.stackexchange, but that isn't really their business model ( http://meta.stackexchange.com/questions/76974/how-can-i-propose-a-new-site ). They do have Open Data ( http://opendata.stackexchange.com/ ), but it might be odd to host mimic questions on a general site.

Ikaro set up a google group previously, but it was a horribly kludgey and not an ideal solution.

I'm just being devil's advocate. If you think the maintenance and support are feasible then any are probably good.

Are we thinking of hosting it on our servers or AWS?

from mimic-website.

I'd thought about proposing a StackExchange for MIMIC, but I'm pretty sure it would be doomed to failure. You need to have a very active userbase to make it out of the beta, as @Franck-Dernoncourt and I recently experienced (http://area51.stackexchange.com/proposals/65426/open-science).

Discourse looks nice, but I'm not sure it would be "easily" maintained. Anything we have to install ourselves runs the risk of becoming outdated and unmaintained (see joomla, redmine).

Everything becomes outdated at some point, so the best option is to choose something that has a strong community and limited tie-in. If you prefer an externally hosted solution, then we could pay a subscription to Discourse (https://payments.discourse.org/buy/) or choose an alternative platform.

from mimic-website.

What platform are you guys most comfortable with? Ruby (Discourse), Django (Askbot, OSQA), or PHP (just kidding)?

Hosted solution is good, except that Discourse is $100/month. I think Askbot was $15/month. Maybe we should install one that we think is good, and if the users like it, then we can justify a hosted solution moving forward.

from mimic-website.

I agree, $100 would be a lot to pay for a small forum, so hosting locally would make sense. As we aren't planning on any major customization, we should be able to avoid getting involved in the codebase, so from that perspective any of the languages would be fine.

from mimic-website.

What about using StackOverflow and just having users tag their questions with something like "MIMIC-ICU"?

from mimic-website.

On StackOverflow questions have to be about some libraries, programming languages, or programming tools, not on data sets.

I was thinking about the open data SE but it doesn't seem to fit there either from my understanding of the on-topic page. I've pinged their meta in case it is though: Is asking questions about a particular data set on-topic?

Also not on-topic on http://health.stackexchange.com/ :-/ (the current moderation there anyway is really not good)

from mimic-website.

Turned out asking questions about a particular data set is on-topic on http://opendata.stackexchange.com/, so it could be used for MIMIC. (not sure if they won't change their mind if they see hundreds of questions though, but since all contents on there is cc by-sa 3.0 anyway, and it's really not common they remove questions due to change of scope). If anyone asks a question about MIMIC there just ping me and I can create the tag (new users can't create tags).

from mimic-website.

@Franck-Dernoncourt Thanks for doing this!

We have been discussing whether or not http://opendata.stackexchange.com/ is a good place for Q&A and haven't managed to reach consensus.

It would be great to have some opinions shared here (@alistairewj @rjstretch @kepaik @erinhong @li-lcp @kpierceHST @rgmark)

Some positives:

• the format is perfect for Q&A.
• many people are familiar with StackExchange
• moving the conversation to a public forum is a clear message that development of code, support etc, is the role of the community
• no technical burden for the lab to support a forum

Some negatives (/questions):

• limited control over the platform. What happens if legitimate questions get marked for deletion? What happens if the accepted answers are weak. What happens if StackOverflow decides to delete the conversations in 3 years time?
• will questions to get lost? Is it simple to set up a notification system so that we are informed when a MIMIC question is posted? Is there a simple way of searching for all MIMIC posts?
• how do we ensure that questions are linked to a specific MIMIC version? e.g. how do we deal with the fact that a new version of MIMIC might have a different itemid dictionary, which could result in older queries failing?

from mimic-website.

Also @tnaumann @spitz-dan-l @mghassem @ikarosilva - please could you add your thoughts on the point above (http://opendata.stackexchange.com/ for MIMIC support)?

from mimic-website.

Just answering a few questions:

limited control over the platform. What happens if legitimate questions get marked for deletion?

Go to meta and refine the scope, if the scope becomes unsustainable, move somewhere else.

What happens if the accepted answers are weak.

Indeed that happens sometimes, leaving a comment is one way around, not perfect though.

What happens if StackOverflow decides to delete the conversations in 3 years time?

Content is cc by-sa 3.0 so one can legally move the content somewhere else.

will questions to get lost?

What do you mean?

Is it simple to set up a notification system so that we are informed when a MIMIC question is posted?

Yes:

Is there a simple way of searching for all MIMIC posts?

yes through searching all questions with a tag mimic tag.

how do we ensure that questions are linked to a specific MIMIC version?
e.g. how do we deal with the fact that a new version of MIMIC might have a different itemid dictionary, which could result in older queries failing?

one tag per version

from mimic-website.

Hey Tom, I like the idea because (like Alistair said) most of the time I'm redirected to a stack overflow question/answer through google anyway.

Franck addressed all the potential negatives, and in general I think the platform makes sense with the push to have a more open/public "ownership" of the data.

from mimic-website.

Thanks guys, I'm sold! @alistairewj sounds happy with StackExchange too, so we'll make a start with setting this up over the next week or so.

Alongside StackExchange for Q&A we need somewhere for people to post bug reports on the data. We plan to create a repository containing the code used to create MIMIC, where people can raise issues and bug reports.

So, the steps are:

• Archive issues at https://github.com/mimic2/v3.0 then delete the repo.
• Create a new public repository containing cleaned "Building-MIMIC" code (the place where bug reports will be posted)
• Post some seed questions on http://opendata.stackexchange.com/, with the "MIMIC-III" tag
• Update the documentation at: http://mimic.physionet.org/community/contributing/ with instructions on how to get support via StackExchange.

from mimic-website.

I think there is a good potential that putting it on SE will have the added benefit of publicity for MIMIC. Also all for offloading management and administration. I was going to suggest running a hosted open source version, but even that would take a lot of time that would be better spent elsewhere.

My concern would be primarily around usability. Finding information should be easy enough, but posting seems like several steps and liable to human error. It would be nice if we could embed a form on our site to automatically create questions on SE automatically tagged.

I wanted to look into it, but it seems like SE is blocking our IPs at the moment...

from mimic-website.

@kepaik I've had some "disagreement" with some SE mods this weekend but I don't think they banned the 18.* :-) SE seems to be working OK but slow on my side of the campus.

from mimic-website.

I'm able to connect now, by switching off the MIT wifi to the LCP intranet.

A direct link such as:
http://opendata.stackexchange.com/?tags=mimic will be nice a clean for listing previous issues, but not sure if a similar method exists for generating questions.

Also, as a "Beta" site, are there concerns it will eventually be deleted?

from mimic-website.

as a "Beta" site, are there concerns it will eventually be deleted?

Yes in theory, in practice once a site has survived its first few weeks, it is rare it gets closed afterward, I would guess < 5%, though I don't have any number on it. (opendata is over two years old but has not much traffic -> http://stackexchange.com/sites#questions)

from mimic-website.

It wouldn't be ideal if the site was closed, but I think this concern is covered by Franck's point below:

Content is cc by-sa 3.0 so one can legally move the content somewhere else.

from mimic-website.

not sure if a similar method exists for generating questions.

Stack Exchange has an API (https://api.stackexchange.com/) with write access, but it would probably be a pain to wrap it in a form for questions and do the same for comments, and answers, etc. (+ if any spammer posts junk that would probably cause some trouble)

from mimic-website.

I've posted the first question on MIMIC there: Gold annotations for diabetic patients in MIMIC-III

from mimic-website.

Thanks @Franck-Dernoncourt -

We'll try to post a response shortly (it'll need some thought!).

from mimic-website.

We are testing Discourse on our internal servers and have a couple of minor issues to sort out. There should be a version ready for public use within a week or two.

from mimic-website.

A test instance of the forum is running at: https://mimic.physionet.org/forum/

from mimic-website.

The forum is currently unreachable because the server hosting it is down.

from mimic-website.

Hi everyone, any updates on the forum? I really feel the need to discuss things with you on a forum or another means. Any chance we can start one?
I know I and other people have many questions and also suggestions and solutions that will help others.

from mimic-website.

@manashty Hi Alireza, thanks for letting us know that there is still demand for a forum. We have no immediate plans to introduce one, due to limited resources, so for now we suggest raising issues on the MIMIC Code Repository when you have points for discussion. We are in the process of updating PhysioNet and we can consider providing a forum as part of this work.

from mimic-website.

A combination of issues and Gitter (https://gitter.im/MIT-LCP/mimic-code) seems to be a good solution for now. We can push more people towards Gitter by linking to it from https://mimic.physionet.org/community/contributing/

from mimic-website.