[Python-Dev] RELEASED Python 3.0 final (original) (raw)

Guido van Rossum guido at python.org
Sat Dec 6 18:54:18 CET 2008

On Fri, Dec 5, 2008 at 9:28 PM, <glyph at divmod.com> wrote:

On 5 Dec, 06:10 pm, guido at python.org wrote:

On Thu, Dec 4, 2008 at 11:27 PM, <glyph at divmod.com> wrote:

With all due respect, for me, "library support" and "serious use" are synonymous.

Glyph, I cannot have a discussion with you if every single post of yours is longer than my combined daily output. Please spend some time writing shorter posts. I'm sure I'm not the only one here with a short attention span. :-) I already spend a lot of time trying to remove extraneous details. The drafts of these messages are usually 3x as long :). So, trying to keep it short:

Thanks!

Thomas paraphrased my point pretty well. The importance of libraries cannot be overemphasized. Maybe you're right and the stdlib is enough for a large audience, but I don't know that audience. Everyone I know who uses Python, uses it because of a library. In some cases, an equivalent library exists for another language, and Python wins because it has a nicer syntax. But, in no case does Python win where it doesn't have the library.

Clearly you're not reading the edu-sig list. :-)

I think that the marketing for py3 needs to target library vendors before targeting novices. If the novices are targeted first, they are going to have a bad experience when "python" libraries don't work with py3, and library maintainers are going to have a bad experience when clueless newbies harass them to update their software without understanding the magnitude of the work to do so.

I think it's great to have specific marketing targeted towards library developers. I know we haven't done enough -- for example I promised a C extension porting guide which didn't materialize. :-(

But I do not think it is a good idea to emphasize elsewhere that most people shouldn't use Python 3.0. Py3k will have a hard enough time gaining mindshare without the very developers who created it discouraging its use. If you can't find it in your heart to recommend 3.0, can you at least keep that within your circle of library-producing friends?

Whenever someone asks me which version to use, I alwasys respond with a question -- what do you want to use it for? And then I'll give them an answer based on what's available for their needs. Sometimes I have to recommend Python 2.2. It's been a while since I had to recommend 1.5.2 but a few years ago that was still common. (A large company I know still has servers where 1.5.2 is the default, although 2.4 is also installed.)

I've been predicting this for years, but two days into Python 3's release, I've already seen real-world examples of this pattern in #twisted. I can tell these people to "downgrade" to py2 when they come ask me for help, but I don't think most of them ask for help. They just get angry and learn Java instead.

If they're that easily convinced that Java is better they probably were a lost cause anyway, so I won't mourn their departure too much.

The one thing I would warn against is replacing a default Python 2.x with Python 3.0 -- if you find 2.x pre-installed, it's likely that other parts of the OS infrastructure depend on it, and any upgrade except to 2.x.n is likely to cause trouble.

-- --Guido van Rossum (home page: http://www.python.org/~guido/)

More information about the Python-Dev mailing list