[Uim] A proposal of unified input method C API.

[YamaKen]

James,

I want to paricipate with the discustion because the API will be
published whether uim deveplers are involved or not.

I suggest that create new dedicated English mailinglist because
many issues are waiting to be discussed. scim list is not
comfortable to host such a lot discussion.

I'll suggest many specification changes such as
following. Whether the API will be used as primary client API
for uim or not, uim cannot adopt the API without such
specification changes.

- License
- ABI-safe extensibility (uim need this)
- Embedded playform capability (i.e. eliminate desktop assumptions)
- Core/Optional feature separation
- Key event specs

I think that creating new project on sf.net such as
unified-im-api to host the mailinglist and header files is
useful and appropriate for the API. How do you think about it?

At Fri, 17 Dec 2004 09:26:46 +0800,
suzhe at tsinghua.org.cn wrote:
> 
> Hi,
>   - Yes the API may be changed in order to add some features or fix some 
> bugs.
>   - There is a maillist for this API, which is on yahoogroup. But it's 
> for Chinese developers only. Anyway, you may discuss this API on scim's 
> maillist.
>   - I'll maintain this API. Actually this API is part of Chinese Linux 
> Standard, and I'm a member of this standard group, in charge of this API.
> 
>   For the feature of surrounding, we considered it some time ago. But 
> because it's hard to implement in some framework, eg. XIM, IIIMF, SCIM 
> etc., and this API is designed for all possible frameworks. So we must 
> ensure that at least the major frameworks can support this API.
> 
> Regards
> James Su
> 
> TOKUNAGA Hiroyuki wrote:
> 
> >On Wed, 15 Dec 2004 16:55:10 +0800
> >James Su <suzhe at tsinghua.org.cn> wrote:
> >
> >  
> >
> >>  Please have a look at the attached header file, it's a standard
> >>  input method C API which is included in Chinese Linux Standard
> >> (I'm a member of it). All members in this standard group have promised
> >> to support such API (including Sun, redhat, Novell/SUSE, ThizLinux,
> >> SW-Linux, etc.)
> >>
> >>  This API does not provide any library, and does not need to link 
> >>against any external libraries. It's just a pure C API which uses 
> >>function pointers to realize the communication between input method 
> >>modules (IME) and input method manager (IMM, eg. uim, scim, etc.). So 
> >>it'll have good binary portability and very easy to deploy.
> >>
> >>  Sorry that there is no english documentation yet, though a chinese
> >>  doc is available.
> >>    
> >>
> >
> >I don't have a time to comment about concrete API for now, but I have
> >some questions.
> >
> >- Is there a plan to change this API?
> >
> >The API of stantard_im.h seems lacking some features. (i.e. surrounding
> >text.) Is adding/modifying to the API possible?
> >
> >- Is there a mailing list or something to disscuss about this API?
> >
> >- Do you have a will to maintain the API? (Or can I maintain?)
> >
> >Of course stable API/ABI is quite important thing, but an evolution of
> >software is also important. Sometimes the evolution of a software need a
> >modification of API. If no one maintain the API, we can't use that with
> >peace of mind.
> >
> >As a conclusion, I think that to make unified input method API is hard,
> >but worth. We should go toward this direction.
> >
> >
> >Regards,

-------------------------------
YamaKen  yamaken at bp.iij4u.or.jp