On 23/04/11 23:16, Matthew Toseland wrote:
> On Saturday 23 Apr 2011 22:51:47 Juiceman wrote:
>> On Sat, Apr 23, 2011 at 1:58 PM, Matthew Toseland 
>> <toad at amphibian.dyndns.org> wrote:
>>> This is a very small and very promising patch, but we would need to
>>> include the JNA library. The JNA library allows us to conveniently call
>>> native APIs without JNI code. We could use this for not only what this
>>> patch does - set background mode to significantly reduce system impact
>>> of our on-startup disk I/O - but also for setting thread priorities on
>>> unix, for which we currently have a native JNI library. And there may
>>> well be other things we'd like to call native libraries for too.
>>> 
>>> Also its platform detection code might well replace our existing code in
>>> that area - or at least some of it.
>>> 
>>> The catch is the library is 900KB, and the platform library (which I
>>> think just has some standard convenience functions) is 700KB. So we
>>> might be looking at increasing the download by 1.6MB.
>>> 
>>> Thoughts? I'm in favour but not totally convinced...
>>> 
>>> The original pull request (more info in attached messages): 
>>> https://github.com/freenet/fred-staging/pull/29
>>> 
>>> 
>> 
>> The ext is updated very rarely so it should not be an issue with the 
>> update script, and users are used to downloading large install files since
>> there is such bloatware out there.  Our main user base has broadband so it
>> shouldn't be a big problem for them.  How will it affect our hosting
>> bandwidth?
> 
> The download is already nearly 10MB. So it's not a huge increase.
>> 
>> Imho, the benefits of improving system responsiveness far outweigh the 
>> minor increase in file size.
> 
> Yes. And we get a reduction in code complexity, maintenance and probably an
> improvement in cross-platform support too (from getting rid of the
> nativethread libraries).
> 

Hosting bandwidth is a non-issue. The main issue is maintainability. Unless
you're prepared to help me maintain freenet-ext, please don't make comments or
proposals about what to do with it.

Adding more and more libraries to freenet-ext in its current form is not a good
idea. This will have to wait at least until after the legacy version is 
deprecated.

Afterwards, yes it's definitely a good idea, but it should only be bundled once
we can use it fully (i.e. port other code like NativeThread to use it too).
It's untidy to bundle a library that's only being half-used, and if we don't
use it fully now we'll forget about it later.

More details on the pull request
https://github.com/freenet/fred-staging/pull/29

X

-- 
GPG: 4096R/5FBBDBCE

Reply via email to