Currently (as of df69e75) it checks only the initial letter (via `mode.startswith`) of "mode" in gzip.GzipFile.__init__. See also: https://github.com/python/cpython/blob/df69e75/Lib/gzip.py#L183

I'd personally suggest that it takes also "+" into consideration, which shall be regarded as writable.

One common (hm.. at least I've observed more than once recently) case where people may make mistakes is to use a tempfile.NamedTemporaryFile as the file object without specifying explicitly a mode, while expecting the GzipFile to be writable.

A quick example (in Python 3.7.11):
```
>>> import tempfile
>>> tempfile.NamedTemporaryFile().mode
'rb+'
>>> import gzip
>>> gzip.GzipFile(fileobj=tempfile.NamedTemporaryFile(mode='rb+')).writable()
False
>>> gzip.GzipFile(fileobj=tempfile.NamedTemporaryFile(mode='wb')).writable()
True
```

Unfortunately this change request may be backward-incompatible, since previously modes like "r+b" have been treated as read-only. See also: https://github.com/python/cpython/blob/df69e75/Lib/test/test_gzip.py#L464

Yes, this is a breaking change, it will break the code that reads gzipped temporary files.

Note also that even if this change be accepted, you could not use this feature until Python 3.9 be released. But you can write to temporary files in all Python versions if specify an explicit mode.

So the benefit is small, but there is a risk of breaking existing code.

There is an opposite proposition: deprecate guessing the GzipFile mode (issue28286).