The cgi module provides undocumented feasibility for logging. cgi.log() formats log message and appends it to the log file with name specified by cgi.logfile if it was not empty before the first use of cgi.log().

One of problems is that it uses the locale encoding for log file. Therefore the result depends on the locale at the moment of the first use of cgi.log().

We can fix this by using some fixed encoding (UTF-8). Or maybe just remove this undocumented feature.

I am for keeping this functionality. Unless others in this nosy list think otherwise.

Which functionality?

- cgi.log()

- opening with current locale

I don't mind keeping the function, but if the file isn't already opened I think using UTF-8 is an appropriate choice.

My bad. I meant cgi.log(), I pressed submit changes in a hurry.

+1 for utf-8.

Available options:

1. Do nothing (keep cgi.log() and continue to use the default encoding for open()).
2. Remove cgi.log(). I think that the deprecation period is not needed because the function is not documented, is not imported by star-import, and is not shown in help.
3. Make cgi.log() using UTF-8 for open(). This may break some existing code if cgi.log() is ever used.
4. Completely rewrite cgi.log() using the logging module.

In all options except 2 cgi.log() needs to be documented and advertised as a new feature. And we should ask ourself: do we need this feature? Does it have advantages over the logging package?

It was more visible in 2.0. But since adding __all__ in 2.1 (in e99d5ea25ba994491c773d9b5872332334ccd1c5) it is a hidden feature.

+1 for removing.
But let's emit DeprecationWarning for now, and remove it later.