Python docs (http://docs.python.org/library/time.html#time.time) say that time.time() function should return UTC timestamp, but actually i get local one:
>>> time.mktime(time.gmtime()), time.time(), time.mktime(time.localtime())
(1313466499.0, 1313480899.384221, 1313480899.0)
As you can see, the result of second statement is equal to result of the third, while it must be equal to result of the first. Checked on 2.7 and 3.1. My OS is Gentoo/Linux, timezone-info is the latest version (2011h).

Maxim Koltsov wrote:
> 
> New submission from Maxim Koltsov <kolmax94@gmail.com>:
> 
> Python docs (http://docs.python.org/library/time.html#time.time) say that time.time() function should return UTC timestamp, but actually i get local one:
>>>> time.mktime(time.gmtime()), time.time(), time.mktime(time.localtime())
> (1313466499.0, 1313480899.384221, 1313480899.0)
> As you can see, the result of second statement is equal to result of the third, while it must be equal to result of the first. Checked on 2.7 and 3.1. My OS is Gentoo/Linux, timezone-info is the latest version (2011h).

The description in the docs is a bit misleading.

time.time() returns the local time in number of seconds since the epoch
(1970-01-01 00:00:0O UTC).

UTC refers to the epoch, not the timezone used by time.time().

Then docs must be fixed. By the way, help(time.time) correctly says about localtime.

Would dropping 'in UTC' from
"""
Return the time as a floating point number expressed in seconds since the epoch, in UTC.
"""
be an acceptable solution?  AFAIK there are no "non-UTC epochs".

Maybe add some words about local timezone?

"""
Return the local time as a floating point number expressed in seconds since the epoch.
"""

Seems OK to me.

> Return the local time as a floating point number
> expressed in seconds since the epoch.

No.  Seconds since the epoch is neither local nor UTC.  It is just an elapsed number of seconds since an agreed upon time called the "epoch".  

I would say: "Return the current time as a floating point number representing the number of seconds elapsed since the epoch."  Suggestions for a shorter formulation that would not change the meaning are welcome.  Maybe "Return the number of seconds elapsed since the epoch as a floating point number."

I patched this one up quickly. One has been changed using belopolsky's recommendation, which already reads well.

New changeset 1559a82a3529 by R David Murray in branch '2.7':
#12758: removing confusing mention of UTC from time.time description
http://hg.python.org/cpython/rev/1559a82a3529

New changeset 5615d6b91b53 by R David Murray in branch '3.2':
#12758: removing confusing mention of UTC from time.time description
http://hg.python.org/cpython/rev/5615d6b91b53

New changeset f18767bb66ba by R David Murray in branch 'default':
Merge #12758: removing confusing mention of UTC from time.time description
http://hg.python.org/cpython/rev/f18767bb66ba

Thanks, Dylan.

Alexander Belopolsky wrote:
> No.  Seconds since the epoch is neither local nor UTC.  It is just
> an elapsed number of seconds since an agreed upon time called the
> "epoch".

This statement just seems wrong. And I have just been confused by the current documentation, hence finding this issue. In what timezone is the "epoch"? It makes a difference. It seems with the current behaviour, the "epoch" is _in the local timezone_. So I reckon the documentation is unclear, because the way I read it, I interpretted it to mean UTC. I think it does need to state "in local time".

However, what I'd really prefer is a new function that returns the seconds since the epoch in UTC.

> It makes a difference. It seems with the current behaviour, the
> "epoch" is _in the local timezone_.

No it isn't. Two different machines:

$ LANG=C date
Wed Jan 16 21:47:03 UTC 2013
$ python -c "import time; print(time.time())"
1358372827.5

$ LANG=C date
Wed Jan 16 22:47:21 CET 2013
$ python -c "import time; print(time.time())"
1358372848.2

time.time() *is* timezone-independent.

Now to your question:

> However, what I'd really prefer is a new function that returns the
> seconds since the epoch in UTC.

>>> epoch = datetime(1970, 1, 1)
>>> (datetime.utcnow() - epoch).total_seconds()
1358372978.448235
>>> time.time()
1358372980.176238

On linux/posix, the epoch is *defined* to be 1970, 1, 1 in UTC.  Python just uses whatever the OS defines the epoch to be, as far as I know.