12

Sometimes I check how many users of mathematica.stackexchange.com there are.
I remember that a few weeks ago there were about 15 thousand and recently I've been surprised seeing that the new users are signed with numbers over 18000.
Let's check this site, the new users therein have numbers slightly over 14000. The registry number of new users included 21 pages with 4 X 9 = 36 new users which is less than 800 << 18000 - 15000.

What is going on?

Let's look at this page, the number of user49115 is 14837 while the next one in the registry is Jazz having the number 15838 i.e 15838 = 14837 + 1001.

I know I can find e.g. how many teachers or students there are, but the number of all users is also interesting.

So what is the reliable number of all users (including unregistered) and of those who are regeistered?

  • Some of them are bots that get destroyed quickly by spam filters, which might explain the difference. – rm -rf Jul 17 '14 at 13:02
  • bots, crocks and frogs – Dr. belisarius Jul 18 '14 at 15:19
  • @belisarius you mean fraugs – Yves Klett Jul 20 '14 at 22:05
  • @rm-rf, it seems there was a different reason. I wonder now if what you are saying is correct, that SE can destroy accounts from bots in the site. – carlosayam Jul 29 '14 at 1:46
  • @caya I don't know what exactly SE does behind the scenes, but I certainly can destroy accounts and have done so. That still won't explain a contiguous block of 1000 users missing, so Oded's answer is the only explanation. – rm -rf Jul 29 '14 at 2:00
9

This is due to how our databases are setup and operate.

The Id field (user number) is an auto incrementing field - essentially when a user gets created, it takes the next number (not entirely accurate, there are some wrinkles there, not relevant to this).

We operate with replication - our databases are replicated between data centers to help with backup and for disaster recovery (if a datacenter/database goes down, we have a replica with near to current data).

In our always-on configuration, to avoid too much network chatter between replicas, seeds are pre-allocated in chunks of 1000 (standard behavior for SQL Server 2012 and above, if I understand things correctly).

If there is a replication hiccup, the result is pretty much what you see.

We did upgrade our database servers around the time of this gap - so, it all fits nicely together...

  • Thanks for explaning the problem. So if I uderstand correctly the registry numbers of new users are not reliable measures of the total number of users? – Artes Jul 26 '14 at 23:44
  • @Artes - they never were. The way identity fields work, they will increment also on a failed attempt - so if current max Id is 111, a failed attempt to create a new user will increment the underlying sequence to 112, a successful creation will increment again, to 113, leaving a hole. This is normal for database operations and is one reason to never use these Ids as meaningful. – Oded Jul 27 '14 at 9:40
4

Now that the technical side of this has been answered I'll attempt to address:

So what is the reliable number of all users (including unregistered) and of those who are registered?

The Stack Exchange User Reputation Leagues page provides some apparently useful data:

Q&A for users of Mathematica (12,957 total users)

It also gives a break-down by reputation which provides a useful way to measure participating members:

Total Reputation
Total Rep*  Users
100,000+    1
50,000+     5
25,000+     17
10,000+     44
5,000+      69
3,000+      95
2,000+      137
1,000+      214
500+        362
200+        728
1+          11,285

Why this page says that we have 1,672 more members than those with 1 reputation point I don't know.

From this data I estimate that we have about one thousand members with a reputation of 100 points, indicating either significant participation or activity on other Stack Exchange sites.

On a personal note I am pleased to see that we have 44 major hitters (10k+) now.

(Update by belisarius follows. User Rep. Leagues as of Feb 10, 2015)

Total Rep*  Users
100,000+    1
50,000+     7
25,000+     20
10,000+     51
5,000+      80
3,000+      116
2,000+      171
1,000+      254
500+        443
200+        898
1+          14,641
  • Looking at the users tab there are about 5000 users with 100 rep or more (of which about 3300 have 101) – Simon Woods Jul 25 '14 at 21:14
  • The discrepancy might be because unregistered users are not counted in certain places. – rm -rf Jul 26 '14 at 0:37
  • Thanks for your answer. – Artes Jul 27 '14 at 0:26

You must log in to answer this question.

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