The future of CSSM and Answers

CSSM is a valuable newsgroup.

Unfortunately it is not possible to embed graphics in CSSM posts, but the noise is a bigger problem:

• Without the possibility to edit existing posts, a correction or improvement needs a new post and leaves junk.
• Spam.
• Double posts caused by the latency after sending a message through the TMW pages.
• Rererepeated questions concerning EVAL and the floating point issues.

The number of posts in CSSM has been growing from 1993 to 2010 a kind of exponential with a bump from 2004 to 2006. Then MATLAB Answers started at the beginning of 2011 (red line), and at the same time the number of CSSM posts broke down:

[EDITED: Diagram updated, 02-Jun-2012, Jan Simon] see About CSSM

I have the impression, that Answers is very near to be CSSM 2.0. Currently the communication follows the fixed question<->answer schema usually, but there have been some votings ( Vote if you want functions in scrips / don't want / don't care ), tutorials about this forum, games and wish lists. These meta-posts, which are actually no questions in the sense of a FAQ, got the highest number of votes.

In consequence I think, that there is an important demand for more general discussions in this forum. How can the interface be improved to fullfill this need? Should Answers grow up to Discussions?

We had a lot of very good ideas in the wish-list for MATLAB Answers - this is the thread with the most answers. The TMW team seems to follow the old and wise strategy not to implement new features hastily. While this is method is very efficient to create stable and reliable MATLAB releases, I'd be glad to see faster advances in this forum.

[EDITED] Related CSSM post: 311689

[Fangjun Jiang]

I like the topic!

[bym]

I am struck by the similarity of the above graph to my retirement account value! Could it be all those financial "quants" that used matlab and asked questions on CSSM are now unemployed?

[Sean de Wolski]

proecsm, my curve too just with a date vector beggining in 2008.

[per isakson]

What hinder us from discussing here at Answers? What type of topics do we want to discuss?

[Jan]

@Per: I do not want to solve homework questions. I do not want to waste so much time with repeating "Please format your code" and "What is your question".

[Daniel Shub]

Is it possible to update this?

[Jan]

@Daniel: I tried it, but I'm not able to find any user statistics anymore. About CSSM offers a meaningless diagram only, which was created in the beginning of March, but it is not clear, which year is meant. The detailed per month and per user statistics are vanished.

When I see such an outdated diagram, I tend to believe, that the corresponding service is dead.

Does anybody know an efficient way to get user statistics for CSSM? I could urlead the news interface on mathworks.com, but there are 253465 pages and it took about 0.8 seconds to get one page.

Answer 1

[Oleg Komarov]

Quoting Jan:

"In consequence I think, that there is an important demand for more general discussions in this forum. How can the interface be improved to fullfill this need? Should Answers grow up to Discussions?"

If I recall correctly TMW gives weight to our opinions and I take the opportunity to renew my suggestion to vote on Wish-list for MATLAB Answer sections if you think that a tutorial section could be useful.

Obviously the idea has gotten wider and I would be happy to vote Jan's proposal on the wish list.

Answer 2

[Walter Roberson]

MATLAB Answers is still in the stage of "We'll make it available and see what evolves out of it". If we want it to be more "Discussions", then we can make it more Discussions. All we have to do is start, and to let it be known that we have changed the scope of what is supported here.

That said:

1. This is still a feature of TMW, and TMW is (I would suppose) not likely to be interested in spending much employee time on matters not related to their business.
2. Thus pure theory discussions that do not tend (overall) to increase the use of TWM products should, at the very least, be largely self-running by community volunteers.
3. Theory discussions that tend to suggest that a competitor's product would be more appropriate would, I imagine, be broadly acceptable to TMW only to the extent that TMW is not really competing in that area and has no real interest in competing in that area. For example, it seems unlikely to me that TMW would decide to go head-to-head on performance versus http://en.wikipedia.org/wiki/Great_Internet_Mersenne_Prime_Search
4. Good theory people who are patient with newcomers, able to write well, do not "talk down" to people having difficulty with concepts, and whom do not get involved in theory flame-wars, are unfortunately not in high supply. It may be difficult to get a "critical mass"
5. Even patient theory people are prone to "triage" topics. If in (say) 4 hours, someone can give good hints to (say) 3 people who show evidence of being able to "take the ball and run it down the field", then that theorist is more likely to do that than to risk spending the 4 hours explaining (say) how to form a matrix determinant to someone whose response might be, "Read my posting again, stupid. I didn't ask for a lecture. If you aren't competent to provide the finished code, then just shut up."

Answer 3

[Fangjun Jiang]

I wouldn't call "slow" old and wise. I wouldn't think it is that hard to implement some of the high demanding wish lists, such as better search engine, mark a question for my own interest, edit comments, etc. The order of "My Answers" is still not fixed. The response from the MATLAB Central team has been disappointing.

[Jan]

Ok, but I solved my keyboard latency problem by buying a faster computer after I grouched for half a year (did I mentioned this already?). So I'm "slow" also ;-)

I tremendously appreciate, that the MATLAB releases are well tested before they are shipped to the customers. Stability and backward compatibility are very important for scientific computing in the field I'm working in. Therefore two major releases per year are fast enough for my demands and slowness is wise.

But a bug in a forum is not serious, because I will cost me some minutes of additional work only - if it is fixed after some days. I see in other forums, that a useful enhancement request can be implemented in several *hours* by a single heroic admin. Of course this means, that changes are tested by the forum users and further updates might be necessary. But as long all problems are fixed inside a day, the reputation and esteem profits from the dynamic improvements.

[Oleg Komarov]

IMO, TMW's attitude is secretive.

Corporate policy? Maybe (or for sure; see also the <http://www.mathworks.com/matlabcentral/newsreader/view_thread/295865#795606 FAQ issue>)

Changing priorities? I can understand that, but nobody will condemn if a forum enhancement is "verbously" postponed after being announced.

[Helen Chen]

Fangjun - We are definitely working on many bug fixes and enhancements for MATLAB Answers, including the sorting concern. As Jan alludes to, all MathWorks development teams follow defined processes to make sure that we deliver a high quality product with each release.

[Walter Roberson]

Helen, from a customer standpoint, about the only worst thing than delaying delivery of a product / feature, is not talking about it.

New Answers features appear suddenly, after an unannounced downtime (_usually_ Wednesday 17:30 EDT), and we are left wondering what changed, and left to test by explore by outselves.

You have internal plans and internal priorities, but you don't talk about them with us. We do have _some_ evidence that you keep customer response in to account, but the features that are actually implemented often do not seem to match the features most in demand. We do not know whether that is because for some internal reason those features are difficult, or if there is a shortage of people available to make such people (Randy is the only person we _know_ works on the code), or if you have "taken the feedback under advisement" but have an internal corporate vision of Answers that makes some features more important than the customers express, or whether the features that are implemented turn out to be building-blocks needed to implement the higher-priority items.

A blog (or Answers post) that discussed (as much as practical without giving away trade secrets) the plans, timing, and successes, would go a long way towards healing the feeling of disconnect.

With the present circumstances, fain would I fathom why the preview threshold could not be reduced by a tweak of a number, a couple of minor tests, and a push out to the distributed servers.

Customers understand technical glitches and understand (reasonable) resource shortages, and understand "shunk-works projects" -- but customers do not understand being left in the dark about what is happening and when, not unless Big Intellectual Property is involved.

[Oleg Komarov]

Can't agree more. A blog is not even necessary, a post would do!

[Sean de Wolski]

like why our answers used to be in chronological order (so we know it once worked) but no longer are?

[Jan]

@Helen: I apreciate that you take part in this discussion.

The defined process to create new MATLAB releases is very important. But this forum is in its infancy and a slow process does not conserve a stable system, but blocks the dynamic growing of the features to the needs of the users. Currently much more agility is needed to increase the usability of this forum.

But I did not pay for this service and we cannot have claims for any features. And MATLAB Answers must be very user-friendly with a high usability. But it could. You took the first step when you started the forum without defining the target group and aim explicitely. Now TMW can take the next steps (95, if I look at Walter's wish-list thread).

Answer 4

[Sean de Wolski]

I do miss some of the long theory (not all of it necessarily related to ML) discussions on CSSM (Walter helps here) and I feel like I learned more from it, though Answers wasn't around for my exponential ML learning curve.

This forum is just so much easier to use (I'd get so aggravated waiting 25 minutes for a post to appear across CSSM; or if you made a typo, waiting 25 minutes, replying to fix the typo, waiting 25 more minutes). We can still have discussions here through comments and post updating, it just doesn't happen very often.

More

I'm also curious what the Answerer demographic will look like in a year here. I just finished my MS degree and will likely be done with MATLAB when my license expires, (unless I can convince/justify to my future employer that it's worthwhile). Others are certainly in the same boat as me and others have slowed down/quit entirely.

Maybe we should begin an obituary thread for retired Answerers?

[Walter Roberson]

I do think there is more room for theory here rather than pure "how-to". Perhaps matters such as algorithms for finding the points of intersection of four spheres were not MATLAB specific, but it was fascinating to me to see the variety of scientific problems that people come up with, and then to see how those problems could be reframed in terms of the libraries and toolboxes that MATLAB provided.

Now, it is true that Answers does tend to get a fair number of one or two line posts asking for the complete code to implement some complex problem, and the only moral qualm I find in me about telling _most_ of those people to RTFM, or Talk To Your Proof, or GIYF (or, yes, sometimes even "Consider getting a job at Burger King instead") is the internal smoothing filter to keep this system professional looking. Doesn't mean I don't *think* those things, just that I've learned to "bite my tongue" and express myself more diplomatically. If you see a posting from me asking "Which research paper do you intend to base your solution on?" it is probably code wording for "You obviously haven't made any effort to find out anything about this topic; please go away and do some studying and get back when you have something meaningful to say."

Some of you might, though, perhaps have noticed that I've been deleting tags along the lines of "not matlab related". If a topic is completely unrelated to things that (at least speculatively if not in theory) could be implemented in MATLAB, then I delete the topic. I haven't hit any "I need any essay on the Rise and Fall of the Roman Empire, stat!" topics, but pure spam is usually fairly obvious. More theoretical questions... well, unless there is a history of the person trying to offload their work for us, I leave alone, under the supposition that once the person has a bit of a handle on the theory, that we might be able to guide them to a MATLAB implementation.

In my opinion, our meta-job as volunteers is not just to point to manual sections, but also to help pave the way to guide people towards choosing MATLAB for their work in the face of alternatives. Thus, if we help transform a theory question in to a "MATLAB is a good choice for solving those kinds of problems", then I think we have done a Good Thing.

Should there be, as someone suggested earlier, a separate "Theory" discussion board? I don't know. Some people would use it for the good, but in my experience at lot of doit4me people do not seem to come close to realizing that their topic is open-ended and subject to trade-offs and so on, so unless there was an editorial button to punt questions from Answers to Theory Discussions, I worry that it might not work out in practice.

[Walter Roberson]

"obituary thread" sounds so final, especially since someone might "unretire" after a time.

"Hall of Fame" sounds more positive.

[John D'Errico]

I hope to un-retire one day, but until then, I'll stick to writing an occasional new FEX contribution.

[Jan]

@John: You can un-retire. But you cannot get out of the hall of fame.

Answer 5

[Walter Roberson]

I stopped posting in CSSM. That accounted for a third of the posts right there ;-)

[Oleg Komarov]

The question is why other people didn't move here? Some of them just gazed and went back.

[Jan]

@Walter: 2 thirds: Your questions belong to the corresponding answers!

@Oleg: I explicitely though of Bruno, Steven and Greg, when I've written this post. I'm going to ask Bruno for his reasons. I hope he is coming back to Answers, if CSSM is drying out. But currently all questions are answered sufficiently already. So I assume Answers is boring for them until more questions appear.

[Oleg Komarov]

I have seen in the beginning that the questions were somewhat more technical and challenging, maybe it's not true anymore.

[Sean de Wolski]

@Oleg: I remember some discussion of how they didn't like the fact that Answers is controlled/edited/enforced by TMW creating a conflict of interest, whereas CSSM is you say, you said it, it's yours etc.

[Helen Chen]

@Walter :-)

@Oleg - That's an interesting observation. I hadn't heard that comment from anyone else on about that shift regarding the technical nature of the content - has anyone else noticed this?

[Oleg Komarov]

I just re-read my comment, the CSSM appeared to be technically more demanding not Answers.

[Sean de Wolski]

@Oleg: agreed.

[Jan]

@Oleg: I agree, that the technical level of the questions (and in consequence of the answers) is decreasing, while the noise is incresing in the last month.

Answer 6

[Jan]

@Walter, Oleg, Fangjun:

Thanks for participating in this thread. I'm still hoping that this forum is under development and future versions will consider the needs of the users. But this hope is getting smaller from month to month. Obviously TMW has other preferences.

Currently it seems, like MATLAB Answers can attract 5 to 15 frequent contributors. This is not a critical mass, which ensures the quality of the answers, if e.g. 8 people are on holidays at the same time or if a question concerns a rarely used toolbox.

Currently the noise (unformatted code, questions without details, missing motivation to read the doc or ask google) exceeds the hard scientific information. The externally hosted pictures are going to be dead links soon such that the corresponding questions loose their meaning - together with their answers.

There has been the idea of letting contributors with more than 1500 points edit the questions. At first this is not realized (at least I do not have the possibility to edit), at second I do not think that the most assidious posters should be responsible for the most tedious job of formatting the code of users, who do not care about the readability of their questions. So I'd prefer an improvement of the guidelines and an implementation of the famous wish-list instead of getting more editors.

[Walter Roberson]

For internal technical reasons (that I don't know), Mathworks has to enable each editor individually and it is apparently not nearly as easy as clicking a checkbox and doing a "save". It took a few weeks for them to enable me. I edit a number of questions each day (especially deleting duplicates.)

If the most assiduous posters, by definition those "Showing great care and perseverance", do not do the editing, then who will? The frequent posters are frequent because they _do_ care.

The guidelines for the Community Editors start by saying that there are no "sins of omission". That is, do only as much as you are comfortable and interested in doing. It is not so much a responsibility as an opportunity.

[Oleg Komarov]

I still hope that timestamps and a tutorial section (with limited permission to write) will be implemented.

I agree about the critical mass and I think that timestamps will definitely attract other people to Answers.

The noise hopefully should be reduced by the tutorial section (always on top), even better if the poster is prompted to read the general rules/tutorial on his first questions.

Editing is an opportunity but the quality of a forum is also expressed by its editors. "With (great) power there must also come — (great) responsibility!"

The "killer" functionality would be the auto-suggestion that some forums implement when typing the title of the post.

[Jan]

@Walter: I know for sure, that _you_ have been _enabled_ long before TMW made you an editor.

Are the guidelines for Community Editors public? Should they be? But I admit, that I miss my own point. I'm simply disappointed, that TMW does not use the opportunity of this forum. I'm afraid that CSSM is drying out, and Answers is growing in very slow motion only. Perhaps the TMW team is working hard on Answers2.0, but today none of the 99 good suggestions of your 7 month old wish-list is implemented. Only creating a list of the last 30 answers of a frequent user has become much more complicated - I have to edit the browser's address line manually, how ugly.

Fortunately admins of other forums are more agile and suggestions for improvements are not only very welcome, but implemented.

[Jan]

@Oleg: The "killer" functionality is the impression, that it will take years to improve the usability.

You do not need to be Mr. Zuckerberg to profit from a forum. TMW has gone the half way already, and now they wait for

Answer 7

[Jan]

Last week I asked for the performance of filter: Answers: 23157.

I hoped, that a contributor, who runs Matlab on a multi-core computer, is interested in the speed of the filter command and runs the code I've posted. On one hand I assumed, that this topic is interesting, because filtering consumes much time and the documentation claims, that the command is multi-threaded since R2007a. On the other hand I thought, that perhaps TMW itself is interested in commenting by question. I'm convinced that the support of multi-core processors is very important in the present and the future.

Daniel and Walter replied, but they could not answer my question - an unexpected exception, btw. My conclusion is, that the audience I can reach here, is currently not large enough. I've read the "How to ask a good question" threads again, but I did not find a better way to motivate other users to spend 5 seconds for running the code and 10 seconds for posting the results.

But I do find dozens of users per week, who spend the time for asking homework questions without any details. Sometimes I wish, there could be a fixed input mask, which contains teh fields: "Type and dimensions of the inputs", "Code - properly formatted as explained [here]", "Results or copy of the complete error message - mention the contents of the failing line explicitely", "Question - including a question mark", "Checkbox: I've read the How-to-ask-a-good-question thread".

[Walter Roberson]

Dang, we do not have the parallel tool box, so I am not accustomed to thinking of our server as being multi-core, so I didn't run the code.

[Daniel Shub]

I agree. We do not have the critical mass to provide good and varied answers to questions which require an understanding beyond the getting started guide. The better UI has allowed Answers to siphon some contributors away from the CSSM, but it has potentially made both forums unworkable. I don't think there is anyplace to get answers to "difficult" questions. I wonder if the appropriate place for your filter question is TMW technical support. It seems like such a waste.

[Jan]

@Daniel: Of course I have contacted the technical support already and they have posted it in the internal dicussion board for the developers. While contacting TMW employees personally has been very very successful every time, a post on the developer board seems to be a message in a bottle. After FILTFILT has been updated in 2011a (2010b?) I gave up to wait for a reply to my concerning question from June 2009 ;-)

I imagine, that the participation of the developers in this forum would act like a magnet to skilled Matlab users. On one hand this would help to reach the critical mass of advanced Matlab users, on the other hand TMW could profit from the discussions directly.

[Walter Roberson]

Developer participation is problematic because of Mathwork's fairly tight policy of not talking about plans or work in progress, or (to a lesser extent) about software internals. "Yes, we know why this is happening, but we cannot tell you why and we cannot give you any idea of when it will be fixed."

[Jan]

@Walter: I've chosen the example FILTER intentionally. Sometimes the user-community can be the *source* of information and instead of telling us the plans, the developers could listen for the problems and ask for suggestions. Example: "FILTER works on multiple threads if the input has >= 8 columns. Does this match your needs? Is there a better choice?" This would not *reveal* plans for the future, but allows for having a plan.

E.g. the multi-threaded SUM was implemented at first with a bug (randomly order of partial results), then 2.5 slower than the single-threaded SUM, see http://www.mathworks.de/matlabcentral/newsreader/view_thread/261678 . I think TMW has difficulties with MT and an open discussion in this forum would be more helpful than silent developers, who are cut from the information behind a wall of policies.