Agreed, there's no need for further work since it's trivial to create a new username. Closing as WORKSFORME.

brian wrote:

I thought this was obvious, but apparently not:

It is trivial to create a new username but it is not trivial to transfer the history, that's
why we had Wikipedia:Changing username in the first place.

It is non-trivial to change the attributions in the history; it may require modifying
hundreds of thousands of stored page revision records for a very actively used
account, so it's not something that's likely to be run automatically in response to a
user-initiated command.

Additionally, interaction with user blocking and other issues need to be considered.
Frequently changing names could be easily abused to make it difficult to nail down
an attacking vandal.

brian wrote:

(In reply to comment #3)

It is non-trivial to change the attributions in the history; it

may require modifying

hundreds of thousands of stored page revision records for a very

actively used

account, so it's not something that's likely to be run

automatically in response to a

user-initiated command.
Additionally, interaction with user blocking and other issues need

to be considered.

Frequently changing names could be easily abused to make it

difficult to nail down

an attacking vandal.

I assumed that the page histories only recorded user IDs to save
space and prevent duplication. Otherwise what's the point of the IDs?

Frequently changing names should be easier to track than frequently
creating new accounts.

richholton wrote:

I think the larger problem lies in the area of signatures, which are based on
usernames (not id's). As user Rholton, that is my public id (User:Rholton) and
my signature ([[User:Rholton]]). This appears on all my talk page posts, votes
for various issues, etc.

The "wiki-way" (or at least the "Wikipedia-way") requires a (currently informal)
web of trust that is based on user identities determined by signature (which are
verifiable by edit-history).

brian wrote:

(In reply to comment #5)

I think the larger problem lies in the area of signatures, which

are based on

usernames (not id's). As user Rholton, that is my public id

(User:Rholton) and

my signature ([[User:Rholton]]). This appears on all my talk page

posts, votes

for various issues, etc.
The "wiki-way" (or at least the "Wikipedia-way") requires a

(currently informal)

web of trust that is based on user identities determined by

signature (which are

verifiable by edit-history).

As I understand it, we had this problem when Wikipedia:Changing
username was being used, so we have nothing to lose by adding the
feature I requested.

This is unless people change their usernames excessively, which we
trust them not to.

artslave wrote:

I agree that we should be able to trust users to do this, the same as creating
new accounts -- advise them that it's hard on the database and not to use it
lightly, remind them that they can change their sig without changing their name,
and punish them if they abuse it. (If it really requires judgment calls, could
it be assigned to admins or bureaucrats or stewards instead?)

Many people want to change name because they choose a username before learning
the norms of the community, and then choose to change within a few weeks of
signing up; would these short requests cause fewer challenges? If the
difficulty is based on the number of records to be changed, perhaps a cutoff
could be assigned. "Simple" requests, less than N records, could go through
immediately, while more database-intensive requests could be assigned to a
queue, to be:

1. run at "off-peak" hours (if we have such a thing)
2. run on an offline server (if that's possible),
3. run during downtime required for other reasons (which still might take

months, but at least would get done eventually) or

4. to languish as they currently do waiting for developer attention (but at

least the simple requests wouldn't clutter the list).

If it's only going to become more and more expensive to change names as our
histories become longer and deeper and more compacted, then perhaps we should
change policy to disallow it, or limit it to accounts of a certain age?

brian wrote:

(In reply to comment #7)

I agree that we should be able to trust users to do this, the same

as creating

new accounts -- advise them that it's hard on the database and not

to use it

lightly, remind them that they can change their sig without

changing their name,

and punish them if they abuse it. (If it really requires judgment

calls, could

it be assigned to admins or bureaucrats or stewards instead?)
Many people want to change name because they choose a username

before learning

the norms of the community, and then choose to change within a few

weeks of

signing up; would these short requests cause fewer challenges?

If the

difficulty is based on the number of records to be changed,

perhaps a cutoff

could be assigned. "Simple" requests, less than N records, could

go through

immediately, while more database-intensive requests could be

assigned to a

queue, to be:

1. run at "off-peak" hours (if we have such a thing)
2. run on an offline server (if that's possible),
3. run during downtime required for other reasons (which still

might take

months, but at least would get done eventually) or

4. to languish as they currently do waiting for developer

attention (but at

least the simple requests wouldn't clutter the list).
If it's only going to become more and more expensive to change

names as our

histories become longer and deeper and more compacted, then

perhaps we should

change policy to disallow it, or limit it to accounts of a certain

age?

This (and all of the comments above) doesn't seem to answer my
question: Why don't histories just record the user IDs, so that we
can just store the username in one place and not worry about
changing the histories?

jeluf wrote:

(In reply to comment #8)

This (and all of the comments above) doesn't seem to answer my
question: Why don't histories just record the user IDs, so that we
can just store the username in one place and not worry about
changing the histories?

Basically, because of IPs and convenience. IPs are logged in the revision, with
user id 0.
To make queries easier, the username is also stored in revisions, together with
the numerical user id. To get all revisions and their authors you just have to
make one query using this design, no special treatment needed for anonymous users.

Yes, other solutions would be possible.

avarab wrote:

See also bug 2523, a Special Page to rename users.

brian wrote:

(In reply to comment #10)

See also bug 2523, a Special Page to rename users.

It appears that if that bug has been fixed, so has this one.

zigger wrote:

(In reply to comment #11)

(In reply to comment #10)

See also bug 2523, a Special Page to rename users.

It appears that if that bug has been fixed, so has this one.

Re-opening this request until the new extension is available on the wikimedia sites.

river wrote:

WONTFIX, because the ability to rename any user via the web interface is too much of
a potential security problem. (think XSS issues).

brian wrote:

(In reply to comment #13)

WONTFIX, because the ability to rename any user via the web interface is too much of
a potential security problem. (think XSS issues).

It's no more of a security problem than creating a new user.

Extensions installed on wikimedia servers.
Contact your bureaucrats to rename users.