i have a little request from you, the users and readers of this group:
please help me spread the word. i think rpyc does not get the
attention it deserves on the internet and python community in
particular; it looks as if the project just doesn't seem to gain
momentum.
i'm monitoring the website traffic using google analytics and the
downloads stats on sourceforge... we had a peak with the release, but
the graphs seem very determined in where they're heading. currently
the site gets ~100 hits and ~20 downloads a day -- and we are only a
week after the release of the new version and the brand-new site. if
this is all the hype rpyc can generate, i feel there's something wrong
here.
over the three years of rpyc, i have heard from dozens of people (both
from commercial companies and private users) how rpyc helped them
solve different problems and how happy they are with it. i know of
many places that have patched their versions internally to suite their
needs. many people have told me that rpyc should be part of the python
stdlib. and many users, so i believe, appreciate the library: rpyc
2.60 did reach 3500 downloads (overall) -- we certainly do have a user
base -- albeit a very quiet one.
what saddens me is that the library hasn't gotten widespread
recognition, and has never developed an active community. only two
people helped with the backporting to python 2.3 and 3.0, and that's
about it. it seems that very few people in the python community are
even aware of the library, or have ever tried using it. so i come here
with a simple request -- help the library gain recognition!
here are some easy things you can do to increase the awareness:
* point people to the site
* drop a link in forums/discussions
* say a few words (like, how rpyc is different than traditional rpcs,
etc.)
* blog about it
* if you know famous bloggers, ask them to make a review of the
library
* tell your programmer friends and coworkers
* post reviews on the web
* click the "I use it" button on ohloh.net (http://www.ohloh.net/ projects/rpyc)
* post links on digg, reddit, etc.
and if you have the time, please contribute back:
* porting to 2.3, 3.0
* help maintain the site (it's a wiki, drop me a line if you wish to
join as a manager)
* create screencasts
* answer questions in this mailing list
* submit bugs (but please also try to debug them yourself before ;-)
* suggest ideas, improvements, requests
* if you write patches for internal use, please share them with the
community, or at least tell us about them.
* work on experimental features (like code migration, distribution)
it's kinda hard for me to see that after all long hours i've put into
it, the project's share is still marginal. i do my best to publicize
the library over the internet, but i am only one person -- your help
is crucial for the success of the project.
I've been a user since before it existed (with pyinvoke).
it's so useful that I keep finding new uses for it.
my own use-cases:
*parallel computing
*running several instances of code that uses globals (this dude used
globals and so I couldnt run several instances of his "class" even
though I had to!)
*testing a distributed system
*bootstraping a distributed system (not same one)
*possibly more (I'm senile)..
in every single use-case it transforms what would had been a p.i.t.a
into something easy and fun.
thanks tomer!
I think one of the possible issues about popularizing this library is that someone needs to support it. I'm not trying to be harsh but having found it earlier this week I'm already ready to walk away because of two things:
• I've sent several mails and patches to the list and had no response. This makes me think that the project is abandonware and either I have to think about forking it myself to use it or accept that patches and releases might be very slow. • There is the multiprocessing module in 2.6. This covers a lot of the same bases as this module and is officially supported.
In short, I want this to work and it to be a great library but the only way that can happen is for a community to be built around it. The only way you get a community is by being interactive and responsive to the members of that community. I know that running an OSS project and maintaining code is not an easy thing to do (been there..) but the fact is that it's down to the project leader to make it happen.
Obviously I could just be unfortunate and be catching Tomer on a week off. I'm just guessing from the number of half-resolved threads that this is not the case though. Shame, I'd love this to work out: I do think its good code but I can't run with something that has no support.
> I've been a user since before it existed (with pyinvoke). > it's so useful that I keep finding new uses for it.
> my own use-cases: > *parallel computing > *running several instances of code that uses globals (this dude used > globals and so I couldnt run several instances of his "class" even > though I had to!) > *testing a distributed system > *bootstraping a distributed system (not same one) > *possibly more (I'm senile)..
> in every single use-case it transforms what would had been a p.i.t.a > into something easy and fun. > thanks tomer!
jamie, the fact i didn't reply to your messages within 3 days does not mean the project is not maintained. i have other things to do as well, in case you wondered.
anyway, since i had collected some changes already, and since what you suggested seems logical (after all, broadcasting can't scale up), i made the relevant patches.
the RegisterServer class is no longer a singleton, and is meant to be subclassed to override on_service_added and on_service_removed. it can work on any udp port, and the Server class is now configurable for registry_ip, registry_port and registry_bcast (boolean).
hope it helps, -tomer
On Fri, Nov 28, 2008 at 21:10, Jamie Kirkpatrick <j...@kirkconsulting.co.uk>wrote:
> I think one of the possible issues about popularizing this library is that > someone needs to support it. I'm not trying to be harsh but having found it > earlier this week I'm already ready to walk away because of two things:
> • I've sent several mails and patches to the list and had no response. > This makes me think that the project is abandonware and either I have to > think about forking it myself to use it or accept that patches and releases > might be very slow. > • There is the multiprocessing module in 2.6. This covers a lot of the > same bases as this module and is officially supported.
> In short, I want this to work and it to be a great library but the only way > that can happen is for a community to be built around it. The only way you > get a community is by being interactive and responsive to the members of > that community. I know that running an OSS project and maintaining code is > not an easy thing to do (been there..) but the fact is that it's down to the > project leader to make it happen.
> Obviously I could just be unfortunate and be catching Tomer on a week off. > I'm just guessing from the number of half-resolved threads that this is not > the case though. Shame, I'd love this to work out: I do think its good code > but I can't run with something that has no support.
>> I've been a user since before it existed (with pyinvoke). >> it's so useful that I keep finding new uses for it.
>> my own use-cases: >> *parallel computing >> *running several instances of code that uses globals (this dude used >> globals and so I couldnt run several instances of his "class" even >> though I had to!) >> *testing a distributed system >> *bootstraping a distributed system (not same one) >> *possibly more (I'm senile)..
>> in every single use-case it transforms what would had been a p.i.t.a >> into something easy and fun. >> thanks tomer!
As I said before, I didn't want to offend, I was only trying to communicate the fact that communities require careful nurturing. In the early stages especially it's always going to be down to the project lead to be there to ensure that this happens. Once more people get on board, obviously it is probably a good thing if more people can be given commit bits and the responsibility can be shared: it clearly makes your life easier if that is the case!
I really only wrote what I did because you had made an impassioned plea to people out there to "spread the word". My points were just to make you think that you have to be really willing to back up that desire with a commitment to be there for people wishing to jump on board, otherwise inevitably people will walk away.
I've used your library for a week now and I think it's awesome. I've contributed patches because I think thats the best way to make things happen, and I'm more than willing to get involved further and try to support your work. I guess perhaps if I just say thank you for putting in this work, maybe that can alleviate the feeling that I'm having a go: I'm not. I just think that perhaps you need to ensure people know that you are there, even if you don't have time to respond properly there and then. It makes people feel safe that they aren't investing time in something that is going nowhere.
Anyway, thanks a lot for taking the time to review the suggestions and patches I have made. I have more questions and thoughts I'd like to post to the list so hopefuly we can get on with dealing with those and making the library even better.
A couple of further thoughts on this subject:
• Have you thought about starting an IRC channel for discussion / support? I think that it's one of the best ways of building a community around a project since it gives people an immediate way to share knowledge and ideas. • Have you thought about giving people commit bits to your repository? Or perhaps if you are not willing to go that far you could look at using some other version control system (git / mercurial etc) that encourage people to fork, and make changes, and allow you to pull those into your trunk? You could distribute the chain of command a little more then but with yourself at the top being the final gateway to getting pathes into the trunk.
Just some thoughts anyway. Be interested to know what you think.
> jamie, the fact i didn't reply to your messages within 3 days does not mean > the > project is not maintained. i have other things to do as well, in case you > wondered.
> anyway, since i had collected some changes already, and since what you > suggested > seems logical (after all, broadcasting can't scale up), i made the relevant > patches.
> the RegisterServer class is no longer a singleton, and is meant to be > subclassed > to override on_service_added and on_service_removed. it can work on any udp > port, > and the Server class is now configurable for registry_ip, registry_port > and registry_bcast (boolean).
> hope it helps, > -tomer
> On Fri, Nov 28, 2008 at 21:10, Jamie Kirkpatrick <j...@kirkconsulting.co.uk > > wrote:
>> I think one of the possible issues about popularizing this library is that >> someone needs to support it. I'm not trying to be harsh but having found it >> earlier this week I'm already ready to walk away because of two things:
>> • I've sent several mails and patches to the list and had no response. >> This makes me think that the project is abandonware and either I have to >> think about forking it myself to use it or accept that patches and releases >> might be very slow. >> • There is the multiprocessing module in 2.6. This covers a lot of the >> same bases as this module and is officially supported.
>> In short, I want this to work and it to be a great library but the only >> way that can happen is for a community to be built around it. The only way >> you get a community is by being interactive and responsive to the members of >> that community. I know that running an OSS project and maintaining code is >> not an easy thing to do (been there..) but the fact is that it's down to the >> project leader to make it happen.
>> Obviously I could just be unfortunate and be catching Tomer on a week >> off. I'm just guessing from the number of half-resolved threads that this >> is not the case though. Shame, I'd love this to work out: I do think its >> good code but I can't run with something that has no support.
>>> I've been a user since before it existed (with pyinvoke). >>> it's so useful that I keep finding new uses for it.
>>> my own use-cases: >>> *parallel computing >>> *running several instances of code that uses globals (this dude used >>> globals and so I couldnt run several instances of his "class" even >>> though I had to!) >>> *testing a distributed system >>> *bootstraping a distributed system (not same one) >>> *possibly more (I'm senile)..
>>> in every single use-case it transforms what would had been a p.i.t.a >>> into something easy and fun. >>> thanks tomer!
The changes you made seem to be really elegantly implemented: thanks for giving it some time. I'm unsure whether or not I can use UDP on the EC2 network but I'll give it a go. If not I'll have to think about subclassing for TCP but obviously UDP is optimal here.
> jamie, the fact i didn't reply to your messages within 3 days does not mean > the > project is not maintained. i have other things to do as well, in case you > wondered.
> anyway, since i had collected some changes already, and since what you > suggested > seems logical (after all, broadcasting can't scale up), i made the relevant > patches.
> the RegisterServer class is no longer a singleton, and is meant to be > subclassed > to override on_service_added and on_service_removed. it can work on any udp > port, > and the Server class is now configurable for registry_ip, registry_port > and registry_bcast (boolean).
> hope it helps, > -tomer
> On Fri, Nov 28, 2008 at 21:10, Jamie Kirkpatrick <j...@kirkconsulting.co.uk > > wrote:
>> I think one of the possible issues about popularizing this library is that >> someone needs to support it. I'm not trying to be harsh but having found it >> earlier this week I'm already ready to walk away because of two things:
>> • I've sent several mails and patches to the list and had no response. >> This makes me think that the project is abandonware and either I have to >> think about forking it myself to use it or accept that patches and releases >> might be very slow. >> • There is the multiprocessing module in 2.6. This covers a lot of the >> same bases as this module and is officially supported.
>> In short, I want this to work and it to be a great library but the only >> way that can happen is for a community to be built around it. The only way >> you get a community is by being interactive and responsive to the members of >> that community. I know that running an OSS project and maintaining code is >> not an easy thing to do (been there..) but the fact is that it's down to the >> project leader to make it happen.
>> Obviously I could just be unfortunate and be catching Tomer on a week >> off. I'm just guessing from the number of half-resolved threads that this >> is not the case though. Shame, I'd love this to work out: I do think its >> good code but I can't run with something that has no support.
>>> I've been a user since before it existed (with pyinvoke). >>> it's so useful that I keep finding new uses for it.
>>> my own use-cases: >>> *parallel computing >>> *running several instances of code that uses globals (this dude used >>> globals and so I couldnt run several instances of his "class" even >>> though I had to!) >>> *testing a distributed system >>> *bootstraping a distributed system (not same one) >>> *possibly more (I'm senile)..
>>> in every single use-case it transforms what would had been a p.i.t.a >>> into something easy and fun. >>> thanks tomer!
On Sun, Nov 30, 2008 at 10:39 PM, Jamie Kirkpatrick
<j...@kirkconsulting.co.uk> wrote: > • Have you thought about starting an IRC channel for discussion / support? > I think that it's one of the best ways of building a community around a > project since it gives people an immediate way to share knowledge and ideas.
+1 on an RPyC IRC channel, I think it would be especially helpful for people just starting out with RPyC (which would be most of the users at this point). I would be happy to sit there most days of the week and answer some questions.
Tal, if you are up for it I'll sit in the room with you. I'm always logged in (though a lot of time afk) and would happily idle there. I prefer freenode...any thoughts?
> On Sun, Nov 30, 2008 at 10:39 PM, Jamie Kirkpatrick > <j...@kirkconsulting.co.uk> wrote: > > • Have you thought about starting an IRC channel for discussion / > support? > > I think that it's one of the best ways of building a community around a > > project since it gives people an immediate way to share knowledge and > ideas.
> +1 on an RPyC IRC channel, I think it would be especially helpful for > people just starting out with RPyC (which would be most of the users > at this point). I would be happy to sit there most days of the week > and answer some questions.
Jamie Kirkpatrick wrote: > Tal, if you are up for it I'll sit in the room with you. I'm always logged > in (though a lot of time afk) and would happily idle there. I prefer > freenode...any thoughts?
I'm certainly up for it, though I should mention (disclaimer...) that I haven't really used RPyC in the past year and haven't used version 3 at all.
I've never set up an IRC channel before, so preferably someone who knows what he's doing should set it up. Freenode certainly seems to be the norm these days for open-source projects.
I'm now idling in #rpyc on freenode.org. I think the process of setting up a channel is as simple as entering a room and ensuring that there is always someone there. I'll check that though but anyway, the channel now exists.
> Jamie Kirkpatrick wrote: > > Tal, if you are up for it I'll sit in the room with you. I'm always > logged > > in (though a lot of time afk) and would happily idle there. I prefer > > freenode...any thoughts?
> I'm certainly up for it, though I should mention (disclaimer...) that > I haven't really used RPyC in the past year and haven't used version 3 > at all.
> I've never set up an IRC channel before, so preferably someone who > knows what he's doing should set it up. Freenode certainly seems to be > the norm these days for open-source projects.
The reason why "pyro python" might be getting more hits on Google
search is because PyRO also stands for Python Robotics. So you should
not get dishearthened by that.
RPyC is great!!!
it should get more traffic, and it will.
Make sure...its atleast as easy to understand and use as Python itself
is. And the rest will be taken care of.
Vishal
On Dec 1, 2:57 pm, "Jamie Kirkpatrick" <j...@kirkconsulting.co.uk>
wrote:
> I'm now idling in #rpyc on freenode.org. I think the process of setting up
> a channel is as simple as entering a room and ensuring that there is always
> someone there. I'll check that though but anyway, the channel now exists.
> > Jamie Kirkpatrick wrote:
> > > Tal, if you are up for it I'll sit in the room with you. I'm always
> > logged
> > > in (though a lot of time afk) and would happily idle there. I prefer
> > > freenode...any thoughts?
> > I'm certainly up for it, though I should mention (disclaimer...) that
> > I haven't really used RPyC in the past year and haven't used version 3
> > at all.
> > I've never set up an IRC channel before, so preferably someone who
> > knows what he's doing should set it up. Freenode certainly seems to be
> > the norm these days for open-source projects.
|
