33
\$\begingroup\$

Some posts can at times have several follow-up questions, sometimes a couple of follow-up questions on the same day.

All this brings us to the question of Can there be too many follow-up questions?

When does the follow-up questions become "too much"? What makes it become too much?

Related: What you may and may not do after receiving answers and How to post a follow-up question?

\$\endgroup\$
  • 7
    \$\begingroup\$ Note: This is not intended as a duplicate of a previous question as this question is to establish a site policy, not a feature request. The previous question also seemed to touch on huge projects and not only pure follow-up questions (A follow-up question is a question which contains code that is a newer version of the code in a previously asked question) \$\endgroup\$ – Simon Forsberg Oct 4 '15 at 19:28
  • \$\begingroup\$ How do I do this on this site: post a question asking for some feedback, choose a hint (among the many) I really like and communicate I am interested in getting a review based on my code with that hint included, not the previous (now outdated) version? Maybe that's the question I should ask? At that point I don't personally see a reason to limit the amount of revisions, provided previous questions are closed and don't steal attention, but that's just me. \$\endgroup\$ – Maroloccio Oct 4 '15 at 20:15
  • 1
    \$\begingroup\$ Another note: This question was inspired because of this question, which was the 5th in the series, but is meant to be applicable to all situations where multiple follow-ups are posted. \$\endgroup\$ – Simon Forsberg Oct 5 '15 at 16:01
  • \$\begingroup\$ @Maroloccio You say "choose a hint (among the many)", but in your case there was mostly only one hint (answer) to choose from. What you should aim at is improving the quality on your follow-up questions, see my answer here for more details. \$\endgroup\$ – Simon Forsberg Oct 5 '15 at 16:03
  • \$\begingroup\$ Simon, all we have is a UI problem, I think. If we had a commit stream like in GitHub, comments could follow the commits and all would be good. Code Review applied to a particular "point in time". \$\endgroup\$ – Maroloccio Oct 5 '15 at 16:31
22
\$\begingroup\$

The problem is not the number of follow-up questions. The problem is the rate of posting them. If you post a follow-up once a week, probably nobody will even notice. As long as it's well done, for example by following the great tips in this other discussion.

Posting follow-ups too soon is not good for anyone:

  • Posting too fast is not good for you. Given enough time, you might receive opposing, conflicting advice on the same question. If you accept a sub-optimal suggestion too fast, you waste your own time. Better to give it a few days, let multiple answers come in, each addressing different angles, and combine them all together once you have a better overview of multiple aspects and considerations. Also, in the example in question, many of the reviews on later posts would have applied to the first, given enough time.

  • Posting too fast is abusing the community. You may end up dominating the front page, and crowd out other questions, which is just not fair.

  • Posting too fast messes with the reputation system:

    • On the one hand, it may seem that you'll get more rep by posting more questions. (I'm not saying this was your intention.) But it doesn't look good, suspicious behavior, and the practice is not to be encouraged.
    • On the other hand, users may be dismayed by this activity and start downvoting. This is not desirable either: posts should be voted up or down on their own merit, not influenced by the behavior of the poster.
  • Posting too fast, too similar questions, can be confusing to reviewers. The thing is, the questions in the example have common parts, which can be "attacked" in multiple revisions. Which revision should I address them? All of them? (REPZ!) It's a mess.

  • Posting too fast, too similar questions, can be boring to reviewers. In the example above, I reviewed rev2 and rev3, and lost interest in reading further. It's just not interesting anymore. It's better to keep the site content interesting.

Lastly, the poster of the example series has received really a lot of attention. Lots of good answers by a high number of distinct reviewers, including many regulars, over a very short period of time. This is clearly not usual, and not normal. Most questions don't enjoy such popularity, and not because they are of lesser quality. The reason this question received so much attention was because it was unusual activity. This incident sets a bad precedent if we don't step up against it. I think it was not intended this way, but the outcome is an abuse of the site.

\$\endgroup\$
  • 3
    \$\begingroup\$ Exactly what part of the site has been "abused"? By my observation, the only abuse was directed at the OP, not by the OP, and he's been nothing but polite and sincere. Also, by the exact same logic, people should not be allowed to answer questions too often. Read your entire post and substitute "answer" instead of "ask", or "question". \$\endgroup\$ – rolfl Oct 5 '15 at 20:03
  • 10
    \$\begingroup\$ I thought the OP took the criticism surprisingly well. I noticed it and I liked it. And I don't think he intended to abuse at all. If you look closely, I even disagree with the downvotes: they are directed at the OP, not the post itself, which is not the intended purpose of the voting system. I thought I made the abuse part quite clear. The repeated questions are crowding out other questions, annoying, and not interesting. What if many users did this regularly? Our site would become pretty crappy. I don't want to see 3+ similar versions of the same question, and definitely not on the same day. \$\endgroup\$ – Stop ongoing harm to Monica Oct 5 '15 at 20:12
15
\$\begingroup\$

The only logical point at which a follow-on process becomes too much, is when the code becomes a duplicate of previously posted code.

Follow-On questions are important on this site, they add a lot of value as they often lead users through a progression of steps resulting in great code.

My experience is that, when a particular user takes their code though a series of follow-on questions, I start to lose interest if there's nothing new to say, so I simply look at other questions, and then, later, see that there's been a new answer, with a different perspective, or insight.

I don't see there being a limit to having "too many follow-on" questions, and, having a 'dangling', unanswered, 'last item' in the chain, is not a long-term problem. Someone will come along and answer it, sometime.

If you are concerned about this as a problem, it may be more because some people who post lots of follow-on questions are also the people who don't apply previous suggestions to their code, so are not improving their code in certain ways between questions.

I find people like that to be frustrating, but there's nothing technically wrong with people who don't follow the advice they are given... I just start ignoring them... if they ignore my suggestions.... and move on.

\$\endgroup\$
  • 3
    \$\begingroup\$ So how do you feel about this question having score -7 ? \$\endgroup\$ – Simon Forsberg Oct 4 '15 at 19:41
  • 1
    \$\begingroup\$ How do I feel about it? Largely I don't care but I am mildly amused by the fact that (so far) scores on previous versions are positive. They should all be negative if the reasoning applied. This is an effect of the question remaining open, I think. The others got asked and answered quickly, so they were upvoted "along the way". \$\endgroup\$ – Maroloccio Oct 4 '15 at 20:21
  • 1
    \$\begingroup\$ @Maroloccio a checkmark is never final, nor does it prevent new answers. As the asker you can revoke a checkmark any time, and mark a new answer as accepted any time. \$\endgroup\$ – Mathieu Guindon Oct 4 '15 at 20:26
  • 1
    \$\begingroup\$ @Maroloccio - I am pleased you are part of this conversation. FYI, I have identified a number of other bugs in your code and I am preparing an answer to your question on the main site. \$\endgroup\$ – rolfl Oct 4 '15 at 20:28
  • \$\begingroup\$ @Mat'sMug: How does that cover acknowledging 2 valid inputs which touch on 2 different algorithmic aspects? At least with the revs you can recognise contributions as you go along... \$\endgroup\$ – Maroloccio Oct 4 '15 at 20:28
  • \$\begingroup\$ @rolfl: great that you keep improving that code. You keep teaching, I keep learning. I think the process works like this, doesn't it? It's not "broken"... Perhaps it could be "better". \$\endgroup\$ – Maroloccio Oct 4 '15 at 20:29
  • 2
    \$\begingroup\$ @Maroloccio that question is about as old as this site, see which answer to accept?. Also relevant: Ĵack of all trades, master of none - team review. \$\endgroup\$ – Mathieu Guindon Oct 5 '15 at 1:43
  • 1
    \$\begingroup\$ @rolfl "(...) is when the code becomes a duplicate of previously posted code" - what is your definition of duplicate here? I believe that you have previously said on meta that only exact duplicates (where code matches 100%) is duplicates. As such, I doubt that will ever happen for follow-ups (then they wouldn't be follow-ups anymore, just duplicates). \$\endgroup\$ – Simon Forsberg Oct 5 '15 at 13:53
  • 1
    \$\begingroup\$ @SimonForsberg - that is exactly my point - that duplicates have to be exact copies, and, as long as the code evolves between questions, there is no technical limit to the cycles. Note that 5-cycles-so-far is hardly excessive. People are anticipating problems that don't, and won't exist. \$\endgroup\$ – rolfl Oct 5 '15 at 15:21
  • \$\begingroup\$ @rolfl In this recent case, the problem is not with the follow-ups themselves, in my opinion the problem was the lack of quality and the short time frame in which the follow-up questions were posted. I have nothing against 100, or even more, follow-ups. But I would require a lot from those questions to not down-vote them. \$\endgroup\$ – Simon Forsberg Oct 5 '15 at 15:55
13
\$\begingroup\$

On the same day? Depends on the substance of the question and of the answers received, but as a general rule of thumb I'd say... one... but that can't be carved in stone: it depends™.

There simply can't be a hard limit.

But flooding the front page with several very similar questions where all but one are marked as answered, is eating up other questions' visibility: the front page can only ever display X many questions.

It's a matter of balance, common sense, and respect of the community: we get more than enough questions daily to get an older version off the front page - waiting at least a whole day between revisions has a number of benefits:

  • It doesn't flood the front page with different revisions of the same code
  • It gets more views on the "current" version, and quite possibly more answers
  • If it's a good question with a positive score, it counts toward the curious, inquisitive and socratic badges; only 1 question per day can count for those.
\$\endgroup\$
12
\$\begingroup\$

Thought I'll chime in here as the other answerer in the downvoted revision 5 question...

To be frank, I was hesitant about answering that revision initially, as I do not know if another answer is going to come in minutes/seconds before mine, followed by a revision 6 in an hour afterwards, which may not include my newer suggestions. In some sense, just like working with library APIs, I don't think it's too wrong to say that we prefer working with a stable question such that people volunteering their advice here knows that it will remain useful and applicable for some time to come, even if that's going to be just "24 hours".

I also think that we as much as we welcome iterative reviews, we will also prefer those where it showcases additional self-improvements from the OP, instead of mere regurgitating of improvements suggested by others. The latter is easy for anyone to see, but the former is the one that is value-adding for both the OP and for the larger community.

Another more technical way to address is to consider each question as a "commit" on your choice of DVCS. This makes sense after all, since the question is supposed to work well to the best of OP's ability and commits are usually done on working code. If your commit style is a per-line basis, then... good for you. Otherwise, treat suggestions from one (iterative) question to another as the basis of a new "commit", then think about how much further it can be improved by oneself to finalize the "commit" before posting the next iteration.

\$\endgroup\$
12
\$\begingroup\$

Can there be too many follow-up questions?

Considering that some follow-up questions can get massively down-voted, I think the answer is Yes, there can be too many follow-ups (at least in a certain amount of time). I say this is based on my observations of community reactions to certain follow-up posts.

Now, I have no intentions of introducing a fixed limit for follow-up questions, but I want to provide my thoughts about what leads to follow-up questions being down-voted, and what question askers can do about it.

A comment on the fifth version of a question said:

Asking 5 reviews of very similar code with very few differences in one day is too much. You aren't giving time for people to answer your questions properly before you ask a new one

This comment has 15 up-votes.

What can question askers do to avoid the down-votes that may come

When making follow-up questions, I think it is important for the question asker, and potentially for the voters, to keep some things in mind.

  • Don't be too quick. It is often better to take a bit of time between posting follow-up questions. Take time to read previous answers, read up on things that the answers have mentioned, then start to read about things related to that, experiment a bit on your own. From the community's perspective, it is probably not that interesting to see five versions of your code in one day (unless you follow some of the advice below extremely well). As also mentioned by @Mat's Mug, there is limited room on the front page. The more follow-up questions you ask in a short period of time, the less visibility other questions get.
  • Is your question a good stand-alone question? How good would the question be if all links to all previous versions of the question, and all other content that is outside the question itself, would die? Don't just link to previous questions for the context, add context inside the new questions also.
  • Put more effort into your question - If you post the same, or very similar, pure-english description in your question over and over again, it will get repetitive. The more follow-ups you make, the more effort you could put into the questions. Find better ways to present what your code is doing. Don't just repeat yourself.
  • Think about your reason for why you want another review. If you have a specific question about a potential improvement, consider: Can that be a possible answer in my previous question as well? If yes, consider adding a bounty to your previous question instead of posting a new one.
  • Make somewhat substantiative changes. If the changes between one iteration and another are somewhat "minor" (for some values of "minor"), your new question is less likely to be interesting. If however you have pushed yourself to make even more changes than the answers have suggested (introduce a new feature perhaps?), then your newer question will be more interesting. Don't just follow what answers say and post a new question, think for yourself.

Is code ever clean enough?

I think the answer to the question Is code ever clean enough? is also Yes. After all, most questions on this site don't result a follow-up. At some point in time, I'd highly recommend just being happy with what you have. After having posted a couple of follow-up questions, consider coding some other things and learn new stuff and look back to your code a couple of months or one year later, I can almost promise you that you will have found ways to improve your own code.

\$\endgroup\$
  • 3
    \$\begingroup\$ +1 for a nice commentry but -2 for not answering the question: How much is too much? - this "answer" says "too much is too much". \$\endgroup\$ – rolfl Oct 5 '15 at 11:17
  • \$\begingroup\$ @rolfl I did not intend to answer the question of "How much is too much?", that is an impossible line to draw as it will depend from time to time. I also never intended to ask a question of "How much is too much", this question is about "Can it be too much?" \$\endgroup\$ – Simon Forsberg Oct 5 '15 at 15:59
  • \$\begingroup\$ The site does not offer a UI to comment on evolving code, like GitHub does (on GitHub, I can add comments to later commits, to a particular line, etc.). As "questions" are the main interface, as if they were articles that should stand on their own, you are building a process to adapt to that limitation. Otherwise it is not only reasonable, but common for developers to comment on a moving commit stream, right? \$\endgroup\$ – Maroloccio Oct 5 '15 at 16:27
1
\$\begingroup\$

Why would a user direct his attention to an earlier rev of a question when that was marked with an accepted answer and linked to an open "follow-up"? Either desist in chasing the update links and choose another question or look at the latest to add there...

Honestly I was told to accept an answer and post a follow-up, it wasn't my idea. I wanted to edit the original question and append code to it. I was told not to.

\$\endgroup\$
  • 8
    \$\begingroup\$ "Closing" a question isn't "marking it as accepted" (a question is usually closed for being off-topic - closed questions cannot be answered). Also, editing the original question and appending new code is precisely the wrong thing to do. Imagine one question with 5 revisions and 5, perhaps 6 or 7 answers, all addressing a different revision -- that would be completely chaotic. You did the right thing by posting a new question. What we're discussing here is at which point the follow-up of a follow-up of a follow-up of a follow-up [...] is, well, beating on a dead horse. \$\endgroup\$ – Mathieu Guindon Oct 4 '15 at 20:13
  • \$\begingroup\$ I used the wrong terminology. I meant "marking it as accepted". Well, curiously enough the question is not a dead horse if somebody can see a way to improve it but those who do not are by definition not in a position to say "there won't be one". Remember: Perhaps the most important principle for the good algorithm designer is to refuse to be content. - Aho, Hopcroft, and Ullman, The Design and Analysis of Computer Algorithms, 1974 \$\endgroup\$ – Maroloccio Oct 4 '15 at 20:18
  • 5
    \$\begingroup\$ I agree, code is never quite "done", and there's always something to improve. Personally I don't think we need to limit the number of follow-ups per se. Working on an answer. \$\endgroup\$ – Mathieu Guindon Oct 4 '15 at 20:20
  • 4
    \$\begingroup\$ I believe this should be the accepted, and high-scoring answer. It is exactly in line with site policy, and is "the right thing". Downvoters are knee-jerking \$\endgroup\$ – rolfl Oct 4 '15 at 22:32
  • 3
    \$\begingroup\$ The argumentation of this answer is not in line with how the SE system should work IMO. Also it demonstrates a fundamental misunderstanding about how the site works. as such it would be bad taste to delete, but worse taste to not downvote from my point of view... \$\endgroup\$ – Vogel612 Oct 4 '15 at 22:46
  • 6
    \$\begingroup\$ I don't really understand the negative reaction your series of questions has received, they look like good questions to me. That said, it is possible you would have got the same answers with just one question, as answerers often build upon other answers (even if one is marked as accepted). \$\endgroup\$ – mjolka Oct 5 '15 at 0:27
  • 3
    \$\begingroup\$ I'm with @mjolka here. I didn't understand the negative reaction when I saw it either. \$\endgroup\$ – RubberDuck Oct 5 '15 at 12:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .