# Should resource recommendations be on-topic?

Today we got our first resource recommendation question (yay!), asking if there are similar resources to COIN-OR and Google's optimization software.

There are a lot of sites that allow recommendation requests like this, and many others that don't. Recommendations are very helpful to some sites and harmful on others.

For Operations Research, I think it could be helpful to have tool recommendations if they're constructive (having a specific set of requirements, a specific goal to meet, etc).

So, should off-topic resource recommendations be considered on-topic or off-topic?

I agree that questions like this should be on topic. But I think we run into murkier waters if questions ask for recommendations, because that is primarily opinion-based.

What I like about the question you linked to is that it asks, "what other open-source solvers are there?" It doesn't ask, "which is the best open-source solver?" In my opinion, questions like that might be off topic, even if they are more narrow. For example, I would say that

Which is the best open-source solver for integer programming?

would be off-topic, but

Which open-source solvers support automatic cut generation for integer programming?

would be on-topic.

Am I splitting hairs too much?

• I disagree. O.R. and optimization are about what's best. Let people answer with their opinion of what's best and why. Make the thread Community Wiki if necessary, so as to keep the SE powers that be happy (enough). Sometimes (in real life work) I like to mention options which are bad, and explain why, because the bad options might otherwise be pursued. Should I just list a bunch of options, including ones I know are bad, and not say anything about that? Or are pros/cons o.k. as long as you don't say what's best – Mark L. Stone May 31 '19 at 2:03
• @MarkL.Stone maybe post this as a separate answer and folks can vote and/or comment accordingly? – LarrySnyder610 May 31 '19 at 2:51
• Yes, I agree with this answer. This is what happens in most of the sites - recommendations are off-topic, but criteria should be specified! – double-beep May 31 '19 at 10:44
• I think your example question "Which is the best open-source solver for integer programming?" should be on-topic. There is the recently-discontinued Mittelmann benchmarks that answers that question. Perhaps people aren't familiar with it. – Edward Lam Jun 2 '19 at 11:38
• @EdwardLam I agree that the topic should be in scope. I'm just saying that "what is the best?" is too vague and calls for subjective opinion. "What is best" in what sense? So I'd be OK with "What's the best solver for parallelization?" or "What's the best solver for automatic cut gen. for IP?" There's still some subjectivity there but IMO the additional criteria (parallelization, cut gen) reduce that considerably. – LarrySnyder610 Jun 2 '19 at 11:45
• Perhaps the question can nclude possible criteria for what''s best. But in any event, the criteria used (considered) can be part of the rationale provided in the answer. For (fictitious) example, I think solver ROBUSTO is the best open source MILP solver because even though it's not the fastest on the Mittelmann benchmark, it's robust as hell. – Mark L. Stone Jun 2 '19 at 12:06
• Please consider that related sites such as cstheory.se allow for :opinion-based" questions in some cases, and they are very popular, such as cstheory.stackexchange.com/q/1168/225 I would strongly recommend not being so heavy-handed about this. – Ryan Williams Jun 2 '19 at 13:10

In my edit to the question Where can I find open source solvers?, I have added the tag . It would be great if someone can review that edit to bring that tag into action. Update: It's been added.

I agree with @LarrySnyder610. Resource-recommendation questions must ensure that answers are objective and not subjective. If they ask for many types of software/methods etc., the tag big-list would also come into mind.

I will use the examples in Mathematics Stack Exchange (partly because I am most active there). The tag online-resources exists and is for

questions that are asking for online resources in mathematics (references, dictionaries, encyclopaedic collections, online tools, videos and lectures, etc), or asking about the mathematics behind tools and technology found on the internet.

This can be easily altered to suit the intentions of Operations Research Stack Exchange.

• Please consider that related sites such as cstheory.se allow for :opinion-based" questions in some cases, and they are very popular, such as cstheory.stackexchange.com/q/1168/225. In fact that question was taken from mathoverflow. – Ryan Williams Jun 2 '19 at 13:10

I think this type of question should be allowed if they are reasonably specific, but the answers should not be opinion based.

If someone asks "Which is the best open-source solver for integer programming for problem xyz?" evidence based answers could be provided in the form of pointing towards appropriate benchmarks.

However I would consider answers based on personal experience (unless backed by evidence, e.g. someone running the OP's model on a few different solvers) off-topic.

on-topic question:

Which is the best open-source solver for this integer programming formulation of the TSP?

off-topic question:

What is the best LP solver?

In this benchmark by Researcher1 on Testset2, Solver3 was significantly faster than Solver 4, 5, and 6. (Link to benchmark)

Solver1 is the best, I have been using it for years

• Please consider that related sites such as cstheory.se allow for :opinion-based" questions in some cases, and they are very popular, such as cstheory.stackexchange.com/q/1168/225 I would strongly recommend not being so heavy-handed about this. – Ryan Williams Jun 2 '19 at 13:12
• @RyanWilliams Personally I think "opinion-based" is OK, but should be supported by evidence or examples. "It's great, I've been using it for years" is not all that helpful. "I found it to be very effective on network design problems" is. Just my 2 cents. – LarrySnyder610 Jun 2 '19 at 13:17

I disagree with the answers by @LarrySnyder610 and @Michael Feldmeier.

Operations Research and optimization are about what's best (under uncertainty), or at least better. Let people answer with their opinion of what's best and why. Make the thread Community Wiki if necessary, so as to keep the Stack Exchange powers that be happy (enough).

Sometimes (in real life work) I like to mention options which are bad, and explain why, because the bad options might otherwise be pursued. Should I just list a bunch of options, including ones I know are bad, and not say anything about that? Or are pros/cons o.k. as long as you don't say what's best? - I think that would be silly

Unlike the actuary profession, Operations Research is not cut and dried, and does not have a formal body of rigid rules for everything. Discretion is allowed and encouraged - that is part of what attracted me to the field, and away from consideration of being an actuary.

In my experience on related Stack Exchange sites, such as Cross Validated/ and Math Overflow, Community Wiki posts with answers about software and other tools and resources addressing particular needs, and with these answers containing people's opinions and rationale, are some of the most useful threads . Unfortunately (in my opinion), there are almost always a couple of law and order types arguing about those threads.

In summary, let people state their opinions and rationale. Make the thread Community Wiki if necessary.

• Agree. In other stackexhanges one gets opinions all of the time, and it's not a big deal, indeed they can be very helpful. The most extremal example (in the space of relevant ones) is probably academia.SE in which most answers are someone's sage opinion on how to handle a tricky academic situation. – Ryan Williams Jun 2 '19 at 13:16