@Werner The SE API gives you 10k requests per day, either per IP, or per ap&user (i.e. per auth token)/day. Overall, a single user can have 1x 10k quota/IP + 5x 10k quotas per auth token + 300 (no API key)/IP per day (day is a window starting from when you first request after no requests for 24 hours). The intent is that each app have at most 10k requests/day.
Each rep history request can give you up to 100 events. How many requests are needed per user will depend on the user. For example, your rep on Meta would require 15 requests. Thus, how long it will take to get all users will depend on what the average number of requests you need per user. Even it it's 1 req/user, then it will take you 15 days, assuming you're using 10k requests per day and have 150k users.
As to how fast you can make requests, there's a hard rate-limit of 30 req/s. My experience with backoff described above is using 10 req/s (1/3 max). Assuming you got no backoffs (unlikely) you could burn through your 10k requests in as little as 5.6 minutes (assuming 30 req/s).
Realistically, no application is going to make 30 network requests a second. Reaching out to the network is a heavy task, and unless you employ serious multithreading and have a processor with lots of physical cores, hitting 30/s is hard.
@Makyen It's definitely an avenue to consider. Many users, across all networks, are infrequent/short-term and therefore contribute very little to the overall site reputation. We have ~7K users with 200+ reputation. That drops to ~200 users with 10K+ reputation.
Regardless, it's the only feasible way to access unearned reputation if I wish to report on such a statistic. It may still be an estimate, since it would exclude private events, but those are typically small (-1 and -2 reputation) compared to public events (+5, +10, +15, ...).
@ArtOfCode That sounds something like a challenge. :-) However, I'd want to find something other than the SE API against which to test, as I find it quite inconvenient when SE shuts down my access.
Frankly, it didn't seem that difficult to at least approach that number (from the POV of initiating the request). I don't recall if I tested prior to implementing rate limits for initiating requests. However, I certainly wasn't checking what actually hit the wire, as opposed to what was just queued somewhere in the protocol stack.
@Makyen Aye, that's where you hit delays. You can call a method to start a request that often fairly easily - but (a) they queue up locally, depending on your bandwidth and network stack, and (b) once they start coming back and you have to process that, you slow down again
So I found someone asking a question who linked to a copy of their credit and debit numbers encrypted with a crappy homebrew encryption algorithm.
I'm a little amazed that it wasn't a bluff.
Is the link something that should be flagged as PII? Should it be flagged even if the poster explicitly says that if anyone can break the code, the card info is theirs?
@JourneymanGeek The worst part is that, even if I couldn't find a cryptographic weakness, the key space is way too small. I only have to try up to 10^9 keys.
Biggest problem is that the description of the algorithm is so vague that I'm trying to wrap my head around it well enough to implement it in C.
Wait nvm, I just realized the card info is his form of a bug bounty.
So it's not PII after all, just a very valuable bounty.
On an unrelated note, why do community ads need to have a score of +6 to show? Why not just show, say, the top few answers? Small sites may have a harder time of reaching +6 than large sites where even a bad ad can get that score.
I understand that not all languages have punctuation to mark the end of a sentence. That was at least one of the problems when they translated to English.
Proposal: Sanskrit Language
The following proposals were closed relatively recently:
Indian Languages
Hindi Language
Sanskrit Language
It seems they have any chances only uniting.
> Are you a human being? We apologize for the confusion, but we can't quite tell if you're a person or a script. Please don't take this personally. Bots and scripts can be remarkably lifelike these days!
@Suvitruf I don't see why. it is in english and doesn't look like something that isn't already answered or of high value. Let's not migrate crap, not even to the russian sites....
@rene author has russian name (: But yeh, probably, it will be closed on SOru too.
user302202
@JayHanlon Love Stack Overflow, appreciated your post on making it more welcoming. Today I was surprised to notice that simple courtesies like "Hi [name of poster here]" are still being stripped out. Thought the plan was to fix that and humanize things a little more?
> Despite Stack Overflow's recent attempts to make itself more welcoming, its help pages still order you not to comment on helpful answers to say "thank you". twitter.com/scottdotjs/status/1057678139210588160
@Mrs.Robinson The implication is true. SO has failed to be welcoming to them because it still hasn't been able to teach them why we don't want 450k more "thank you"s in posts and 50k fewer accepted answers
When your goal is understanding Feynmann's lectures, it's not that the pen you will use to take notes 'doesn't matter'. Of course, you could say such a thing in the sense that you can simply replace a broken pen.
Well, we got side tracked. Fewer accepts would be a minor side effect of allowing "good anser, much wow" comments.
The main reason is I don't give a damn about PeakyBlinders99 thinking the post was informative when there's already a score besides the post when I most need to read it.
If the fifth answer to a real, legitimate question you had on SO had all the extra clutter as the previous four, there's a significant chance you won't feel like reading the sixth for your sixth legit question.
Basically, it's the conflicts of the two "personal help desk" versus "big library" mindsets, again
Ideally, one thread shouldn't be tailored to the asker's needs or feelings any more than they already do it by the way they're wording their question in the first place. One could see it as robotization or whatever, but by the same logic any reference material would be too robotic because it aims to be applicable to more than just one individual
The limit also applies to IP addresses, so if one's posting from a shared network, they can be blocked by other users as well. — Sonic the Introverted Hedgehog5 mins ago
Haha Sonic is like the trivia section in a Wikipedia article
@Bookends The total number of users in the database, using a simple query, is 9969791, that's 30,209 less than 10M; and it goes without saying that not everyone is active. If you find a better query post a link.
The latest ID is around 11006900, so... over 1 million users nuked? (congrats?)
oh, also account merging...
user302202
11:23 PM
@Bookends Back when abandoned accounts were automatically deleted, the deletion rate was 1000 accounts/day which is a lot more than any merges the mods might do.
Proposal: Sanskrit Language
The following proposals were closed relatively recently:
Indian Languages
Hindi Language
Sanskrit Language
It seems they have any chances only uniting.
I noticed SO just had over 10 million users. Not sure where to shout "hey guys, have you seen, it's 10 million now!".
