On 23 Feb 2008, at 20:23, A.D.F. wrote:
> "A.D.F." wrote:
>> Alvaro Lopez Ortega wrote:
>>> On 23 Feb 2008, at 14:16, Miguel Angel Ajo Pelayo wrote:
>>>
>>>> What do you think about that idea?
>>>>
>>>> 0.6.0 makes cherokee look like an immature software, and stops many
>>>> people from using it.
>>>>
>>>> A 1.0.0 would describe more clearly that it is a mature piece of
>>>> software (it really is), although it wouldn't
>>>> have all the features planned for the original 1.0.0 which are not
>>>> necessary for everyone.
>>
>> I think that 0.6.x should be a test series before doing other
>> important improvements to Cherokee, so from my point of view
>> it is good that Cherokee hasn't gained version 0.9 or 1.0 yet.
>>
>>> It is a very interesting suggestion, actually. In fact, it kind of
>>> makes sense to me.
>>>
>>> On the one hand, bumping to higher release number would change the
>>> perception of some people, and that is a good thing. On the other
>>> hand, it would be kind of weird to introduce the change, and that  
>>> does
>>> not sound like a smart thing to do.
>>>
>>> There might be a third option though. What if we release trunk as
>>> something 'done but not quite dusted' (thinking of cherokee-admin)?
>>> For instance, we release trunk next week as 0.99.0, and we focus on
>>> maintaining it for a few weeks before we branch it to stable.  
>>> Then, we
>>> could release it as 1.0 and start working on - let's call it -
>>> 'Cherokee next'.
>>>
>>> What would you think of that?
>>
>> I think that Cherokee needs at least other two major versions
>> before thinking to switch to 1.0.
>> --------------------------------
>>
>> So, I would keep the 0.6.x numeration, trying to release 0.6.0
>> as soon as possible (i.e. within a week or so);
>> 0.6.x could be maintained for a while (i.e. 5-6 months);
>> every improvement written for trunk/admin could be backported
>> from trunk to 0.6.x as is, whereas only severe bugs (security, etc.)
>> should be applied to 0.6 branch.
>>
>> Within 6 or 7 months we could release version 0.9
>> and then after more tests, when 0.9.x is declared really stable
>> and stunning we could refine all the other parts (i.e.  
>> documentation, etc.)
>> and switch to 1.0.0 stable.
>
> If you want to mark the distance between 0.5.x and current trunk,
> next branch could be named 0.8 (after relicensing it),
> so that in 0.9 we could add some of the planned steps for old 0.7 TODO
> along with new ones:
>
> [ ] Dtrace hooks
> [ ] priv.h on Solaris
> [ ] New header entry
> [ ] Generic caching (HTTP-Cache)
> [ ] Mime encodings separated from mime types
> [ ] Chuncked encoding
>
> [ ] flexible internal error reporting to syslog too, etc;
> [ ] full signal handling;
> [ ] HTTP/1.0 speed up;
> [ ] HTTP/1.1 pipelining speed up;
> [ ] sendfilev() support
> [ ] more I/O parameters;
> [ ] sophisticated file cacheing.
>
> The following ones could be added after 1.0 is released,
> i.e. in 1.1 or 1.2, unless someone contributes and
> donates the source code within the end of 2008:
>
> [ ] X-Sendfile
> [ ] WSGI handler
> [ ] AJPv13 handler
> [ ] WebDAV handler
> [ ] Upload progress module
> [ ] mod_evasive
> [ ] Memcached support
> [ ] Language support (handler?)
>
> The following ones could be added in 2.0:
>
> [ ] AIO based fdpoll
> [ ] splice() support

It does also make sense to me. However, I would reduce the new feature  
number in 0.8 and 0.9 to de minimum. In my opinion, a good roadmap  
would be:

0.8.0
-----
[ ] New header entry
[ ] Mime encodings separated from mime types
[ ] Chuncked encoding

0.9.0
-----
[ ] full signal handling
[ ] flexible internal error reporting (to syslog?)
[ ] HTTP/1.0 speed up (not high priority)
[ ] HTTP/1.1 pipelining speed up (not high priority)

I would postpone the rest to 1.0 or later, actually. In fact, these  
improvements should not take us much time and we could release 1.0  
within... a couple of months?

I think that bumping to a 1.0 release as soon as possible is a good  
thing. But, even more important than reaching the 1.0 milestone soon,  
is to release new versions often.

--
Greetings, alo.
http://www.alobbs.com/





_______________________________________________
Cherokee mailing list
[email protected]
http://cherokee-project.com/cgi-bin/mailman/listinfo/cherokee

Reply via email to