Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[Peter Šurda]

On Wed, 15 Feb 2017 19:33:05, ng0 wrote:
> PyBitmessage doesn't build from git commits here. I nfact, there
> wasn't much public activity in upstream in the last months since the
> last release. We use the tarballs.
PyBitmessage is being developed in the v0.6 branch, that's where you'll
find the commits. The previous comment regarding new dependency on
msgpack was correct. v0.6.2 is planned on being released by the end of
the month. I'll be happy to assist but I'm not on this list, I just
monitor the web for mentions of "bitmessage". For discussing development
of Bitmessage, I recommend using one of the following, in this order:

- The "bitmessage" chan on bitmessage
- https://bitmessage.slack.org
- #bitmessage on freenode

For easier resolving of dependencies, the latest code can also detect
more accurately what sort of OS (including several linux distributions
and BSD variants) it's running on and makes a recommendation for what to
do with your package manager based on that. I don't have GNU Guix in my
test environment now. I can add it but may need assistance as I'm not
familiar with it.

You can also submit a pull request, here are the contribution
guidelines:

https://bitmessage.org/wiki/Contribute

-- 
Peter Šurda
Bitmessage core developer

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[Peter Šurda]

(crossposting to the mailing list and participants)

I managed to get Guix running in a VM and updated the instructions
on the bitmessage wiki for Guix:
https://bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies

(last row in the table). Based on this, whoever maintains the package
should be able to update it.

I haven't actually tested OpenCL because building pyopencl requires
opencl libraries and I have no idea how to install them on Guix. I don't
even think that open source drivers are compatible with the current CL
kernel, I was only ever able to get it running on closed source drivers
(AMD/Nvidia/Intel).

PyBitmessage is now compatible with both openssl 1.0.x and 1.1.x
(actually it should work back until 0.9.8b but I haven't tested it), and
in the next days I'll merge LibreSSL support so if you have to choose a
particular version or something, it doesn't matter, it will work with
whatever is installed.

On Fri, Feb 17, 2017 at 12:05:02PM +0100, Peter Šurda wrote:
>On Wed, 15 Feb 2017 19:33:05, ng0 wrote:
>>PyBitmessage doesn't build from git commits here. I nfact, there wasn't
>>much public activity in upstream in the last months since the last
>>release. We use the tarballs.

-- 
Peter Šurda
Bitmessage core developer

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[ng0]

I don't really feel part of this conversation, but since you keep CC'ing
me I have to reply.

On 17-02-17 17:25:34, Peter Šurda wrote:
> (crossposting to the mailing list and participants)
> 
> I managed to get Guix running in a VM and updated the instructions
> on the bitmessage wiki for Guix:
> https://bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies

This is not accesible from behind restrictive proxies and networks like
tor, so I have to pass on this.
 
I did the initial packaging of PyBitmessage on GuixSD but there is no
maintainer per package, whoever wants to update can update it.

Since all my questions, improvements, and attempts to contribute have
been blocked and/or forgotten by the PyBitmessage project I don't want
to have this discussion in my inbox longer than necessary.

LibreSSL support is at least one improvement I welcome.
Any news on the use of setuptools, and the move to PyQt5?

> (last row in the table). Based on this, whoever maintains the package
> should be able to update it.
> 
> I haven't actually tested OpenCL because building pyopencl requires
> opencl libraries and I have no idea how to install them on Guix. I don't
> even think that open source drivers are compatible with the current CL
> kernel, I was only ever able to get it running on closed source drivers
> (AMD/Nvidia/Intel).
> 
> PyBitmessage is now compatible with both openssl 1.0.x and 1.1.x
> (actually it should work back until 0.9.8b but I haven't tested it), and
> in the next days I'll merge LibreSSL support so if you have to choose a
> particular version or something, it doesn't matter, it will work with
> whatever is installed.
> 
> On Fri, Feb 17, 2017 at 12:05:02PM +0100, Peter Šurda wrote:
> > On Wed, 15 Feb 2017 19:33:05, ng0 wrote:
> > > PyBitmessage doesn't build from git commits here. I nfact, there wasn't
> > > much public activity in upstream in the last months since the last
> > > release. We use the tarballs.
> 
> -- 
> Peter Šurda
> Bitmessage core developer

-- 
ng0 -- https://www.inventati.org/patternsinthechaos/

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[Peter Šurda]

Hello,

On Fri, Feb 17, 2017 at 04:41:31PM +0000, ng0 wrote:
>I don't really feel part of this conversation, but since you keep CC'ing
>me I have to reply.
I keep CCing you because I don't know if my messages will make it into
the list, and I don't know who is doing what.

>> https://bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies
>This is not accesible from behind restrictive proxies and networks like
>tor, so I have to pass on this.
There is at the moment a snapshot mirror available for testing new web
infrastructure with the goal to improve the experience. I manually
copied over the Guix instructions so that you can view those too:

https://test.bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies

>I did the initial packaging of PyBitmessage on GuixSD but there is no
>maintainer per package, whoever wants to update can update it.
Ok.

>Since all my questions, improvements, and attempts to contribute have
>been blocked and/or forgotten by the PyBitmessage project I don't want
>to have this discussion in my inbox longer than necessary.
I'm sorry you feel that way. I'm now doing most of the work and have to
balance between the various aspects of the whole project of which coding
is just one part. I cannot reprioritise based on or even track the
communication of a single user. I need to think about the bigger picture
and whole community. I just bought a 3400 EUR server for VMs so that I
can test the PyBitmessage on the zillions of OS variants around there. I
monitor the web so that I can react to bug reports no matter when they
appear.

The project now has clear(er) contribution guidelines:

https://bitmessage.org/wiki/Contribute

or on the test mirror:

https://test.bitmessage.org/wiki/Contribute

and you can get paid in Bitcoin for providing commits that get merged.

Translations have been integrated with transifex and a webhook will
automatically create a pull request when a language is complete. All
commits are PGP signed.

The current cycle focused on bugfixes and refactoring of some horrible
parts. There are also some minor but nice usability improvements that
are particularly important for beginners.

Extended encoding is available for experimenting and is written in a
developer-friendly way. The project registered #bitmessage on freenode
and the website will be migrated to better infrastructuer after 0.6.2
comes out.

>LibreSSL support is at least one improvement I welcome.  Any news on
>the use of setuptools, and the move to PyQt5?
PyQt5 is not a priority. It may be better to have separate projects for
a library written in C, and a GUI written in, perhaps, PyQt5. But I
don't have time for it, maybe if I can raise funds and pay someone to do
that while I focus on the bugfixes and protocol.

I'll look at setuptools in the current cycle (i.e. in the next week or
two).

>ng0 -- https://www.inventati.org/patternsinthechaos/

-- 
Peter Šurda
Bitmessage core developer

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[ng0]

Hi,

I'm not sure if this discussion is too offtopic for the guix-devel
mailinglist. If anyone thinks it is, please respond.

On 17-02-17 18:19:24, Peter Šurda wrote:
> Hello,
> 
> On Fri, Feb 17, 2017 at 04:41:31PM +0000, ng0 wrote:
> > I don't really feel part of this conversation, but since you keep CC'ing
> > me I have to reply.
> I keep CCing you because I don't know if my messages will make it into
> the list, and I don't know who is doing what.
> 
> > > https://bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies

Interesting, how did you achieve the redirection to .onion? Is the what I guess
to be javascript available publicly somewhere?

> > This is not accesible from behind restrictive proxies and networks like
> > tor, so I have to pass on this.
> There is at the moment a snapshot mirror available for testing new web
> infrastructure with the goal to improve the experience. I manually
> copied over the Guix instructions so that you can view those too:
> 
> https://test.bitmessage.org/wiki/Compiling_instructions#Resolve_dependencies

I think this approach is wrong, or we fail to communicate directly.
If you want to compile PyBitmessage from source
on Guix. there is a "guix package" function for that (refer to the Guix
documentation, invoking guix package). For the target group 'novice
users' you made, a simple 'guix package -i pybitmessage' will do it.
You can instruct this to be run entirely without binary substitutes.

The tests for this package run on our infrastructure, so failing builds
will be picked up for multiple architectures.

I'm not sure wether this page addresses developers or users (that's the
only distinction I would make as documentation and interest of those
differs).

> > I did the initial packaging of PyBitmessage on GuixSD but there is no
> > maintainer per package, whoever wants to update can update it.
> Ok.
> 
> > Since all my questions, improvements, and attempts to contribute have
> > been blocked and/or forgotten by the PyBitmessage project I don't want
> > to have this discussion in my inbox longer than necessary.
> I'm sorry you feel that way. I'm now doing most of the work and have to
> balance between the various aspects of the whole project of which coding
> is just one part. I cannot reprioritise based on or even track the
> communication of a single user. I need to think about the bigger picture
> and whole community. I just bought a 3400 EUR server for VMs so that I
> can test the PyBitmessage on the zillions of OS variants around there. I
> monitor the web so that I can react to bug reports no matter when they
> appear.

My criticism and suggestions were prior to this situation, but thanks
for explaining to me why it has never really been picked up.

> The project now has clear(er) contribution guidelines:
> 
> https://bitmessage.org/wiki/Contribute
> 
> or on the test mirror:
> 
> https://test.bitmessage.org/wiki/Contribute
> 
> and you can get paid in Bitcoin for providing commits that get merged.
> 
> Translations have been integrated with transifex and a webhook will
> automatically create a pull request when a language is complete. All
> commits are PGP signed.
> 
> The current cycle focused on bugfixes and refactoring of some horrible
> parts. There are also some minor but nice usability improvements that
> are particularly important for beginners.
> 
> Extended encoding is available for experimenting and is written in a
> developer-friendly way. The project registered #bitmessage on freenode
> and the website will be migrated to better infrastructuer after 0.6.2
> comes out.
> 
> > LibreSSL support is at least one improvement I welcome.  Any news on
> > the use of setuptools, and the move to PyQt5?
> PyQt5 is not a priority. It may be better to have separate projects for
> a library written in C, and a GUI written in, perhaps, PyQt5. But I

Okay, sounds reasonable. Last year I explained why I'm not really happy
with PyQt4 and why I think depending on Windows XP is a mistake. PyQt5
is available for Windows XP, in case that's the blocker someone failed
to see.

> don't have time for it, maybe if I can raise funds and pay someone to do
> that while I focus on the bugfixes and protocol.
> 
> I'll look at setuptools in the current cycle (i.e. in the next week or
> two).
> 
> > ng0 -- https://www.inventati.org/patternsinthechaos/
> 
> -- 
> Peter Šurda
> Bitmessage core developer

-- 
ng0 -- https://www.inventati.org/patternsinthechaos/

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[Peter Šurda]

Hello,

On Fri, Feb 17, 2017 at 05:53:05PM +0000, ng0 wrote:
>I'm not sure if this discussion is too offtopic for the guix-devel
>mailinglist. If anyone thinks it is, please respond.
I think my emails are moderated anyway because I'm not on the list, so
I'll leave the decision up to the moderator.

>Interesting, how did you achieve the redirection to .onion? Is the what I guess
>to be javascript available publicly somewhere?
I use mod_rewrite to extract the client IP from the cloudflare headers,
match it against a list of tor exit nodes and conditionally redirect it
to the onion address. Then I have a cron job that updates the tor exit
node list from the tor-project website. No javascript.

>I think this approach is wrong, or we fail to communicate directly.  If
>you want to compile PyBitmessage from source on Guix. there is a "guix
>package" function for that (refer to the Guix documentation, invoking
>guix package). For the target group 'novice users' you made, a simple
>'guix package -i pybitmessage' will do it.  You can instruct this to be
>run entirely without binary substitutes.
This is a valid point, having a package is the logical next step.
However that requires me to research how guix packaging works, test and
debug it. And there may be other prequisites for building that I'm not
aware of yet. I don't see that as a productive use of my time. Someone
who already understands how it works however can probably easily get the
list of dependencies from the wiki and create the package in a couple of
minutes. I already have to make sure that there are building
instructions that work for at least 4 other distros, 32bit and 64bit
windows (different), osx, freebsd and openbsd.

>The tests for this package run on our infrastructure, so failing builds
>will be picked up for multiple architectures.
Great. I want to increase the level of automated testing for
pybitmessage as well, that's one of the reasons why I have the new
server.

>I'm not sure wether this page addresses developers or users (that's the
>only distinction I would make as documentation and interest of those
>differs).
Again that's a good point. I would prefer if people with specialisations
contributed to the project instead of me having to learn and do
everything. So I'm trying to reduce the hurdles for developers to
contribute. And maybe someone who specialises in technical writing can
update the wiki.

>My criticism and suggestions were prior to this situation, but thanks
>for explaining to me why it has never really been picked up.
I appreciate your criticism, there was indeed a time in the past where
the project stagnated because basically noone had time, so I decided to
dedicate all my energy to it and I work on it full time now. I hope the
developer community gradually improves.

>Okay, sounds reasonable. Last year I explained why I'm not really happy
>with PyQt4 and why I think depending on Windows XP is a mistake. PyQt5
>is available for Windows XP, in case that's the blocker someone failed
>to see.
At least from my point of view XP compatibility is not the reason for
postponing a move to PyQt5.

-- 
Peter Šurda
Bitmessage core developer

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[ng0]

On 17-02-17 20:33:22, Peter Šurda wrote:
> Hello,
> 
> On Fri, Feb 17, 2017 at 05:53:05PM +0000, ng0 wrote:
> > I'm not sure if this discussion is too offtopic for the guix-devel
> > mailinglist. If anyone thinks it is, please respond.
> I think my emails are moderated anyway because I'm not on the list, so
> I'll leave the decision up to the moderator.
> 
> > Interesting, how did you achieve the redirection to .onion? Is the what I guess
> > to be javascript available publicly somewhere?
> I use mod_rewrite to extract the client IP from the cloudflare headers,
> match it against a list of tor exit nodes and conditionally redirect it
> to the onion address. Then I have a cron job that updates the tor exit
> node list from the tor-project website. No javascript.
> 
> > I think this approach is wrong, or we fail to communicate directly.  If
> > you want to compile PyBitmessage from source on Guix. there is a "guix
> > package" function for that (refer to the Guix documentation, invoking
> > guix package). For the target group 'novice users' you made, a simple
> > 'guix package -i pybitmessage' will do it.  You can instruct this to be
> > run entirely without binary substitutes.
> This is a valid point, having a package is the logical next step.
> However that requires me to research how guix packaging works, test and
> debug it. And there may be other prequisites for building that I'm not
> aware of yet. I don't see that as a productive use of my time. Someone
> who already understands how it works however can probably easily get the
> list of dependencies from the wiki and create the package in a couple of
> minutes. I already have to make sure that there are building
> instructions that work for at least 4 other distros, 32bit and 64bit
> windows (different), osx, freebsd and openbsd.

I think you are trying to do to much, to be honest. Leave the job of
system integration up to the people who are familar with the systems.
We have the guix package already, no point in trying to re-invent it.
Gentoo packagers know best what to do for Gentoo, same for Debian, the
BSD folks, etc. No idea about Windows and OSX.
What I tried to communicate here was, the instructions might be wrong
and misleading. You didn't respond to my question, so I'll just leave
you with this.

> > The tests for this package run on our infrastructure, so failing builds
> > will be picked up for multiple architectures.
> Great. I want to increase the level of automated testing for
> pybitmessage as well, that's one of the reasons why I have the new
> server.
> 
> > I'm not sure wether this page addresses developers or users (that's the
> > only distinction I would make as documentation and interest of those
> > differs).
> Again that's a good point. I would prefer if people with specialisations
> contributed to the project instead of me having to learn and do
> everything. So I'm trying to reduce the hurdles for developers to
> contribute. And maybe someone who specialises in technical writing can
> update the wiki.
> 
> > My criticism and suggestions were prior to this situation, but thanks
> > for explaining to me why it has never really been picked up.
> I appreciate your criticism, there was indeed a time in the past where
> the project stagnated because basically noone had time, so I decided to
> dedicate all my energy to it and I work on it full time now. I hope the
> developer community gradually improves.
> 
> > Okay, sounds reasonable. Last year I explained why I'm not really happy
> > with PyQt4 and why I think depending on Windows XP is a mistake. PyQt5
> > is available for Windows XP, in case that's the blocker someone failed
> > to see.
> At least from my point of view XP compatibility is not the reason for
> postponing a move to PyQt5.
> -- 
> Peter Šurda
> Bitmessage core developer

-- 
ng0 -- https://www.inventati.org/patternsinthechaos/

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[Peter Šurda]

On Fri, Feb 17, 2017 at 07:40:26PM +0000, ng0 wrote:
>I think you are trying to do to much, to be honest. Leave the job of
>system integration up to the people who are familar with the systems.
Based on my experience, without instructions, they often don't do it
correctly. They miss some of the dependencies which then results in
disabled features and then I have to deal with that. That's the purpose
of the wiki page linked earlier.

>We have the guix package already, no point in trying to re-invent it.
>Gentoo packagers know best what to do for Gentoo, same for Debian, the
>BSD folks, etc. No idea about Windows and OSX.
>What I tried to communicate here was, the instructions might be wrong
>and misleading. You didn't respond to my question, so I'll just leave
>you with this.
I guess I don't understand what you're looking for then.

-- 
Peter Šurda
Bitmessage core developer

Re: [PATCH] gnu: Add pybitmessage, [PATCH] gnu: Add python2-pyqt-4, [PAT

[ng0]

On 17-02-17 21:06:27, Peter Šurda wrote:
> On Fri, Feb 17, 2017 at 07:40:26PM +0000, ng0 wrote:
> > I think you are trying to do to much, to be honest. Leave the job of
> > system integration up to the people who are familar with the systems.
> Based on my experience, without instructions, they often don't do it
> correctly. They miss some of the dependencies which then results in
> disabled features and then I have to deal with that. That's the purpose
> of the wiki page linked earlier.
> 
> > We have the guix package already, no point in trying to re-invent it.
> > Gentoo packagers know best what to do for Gentoo, same for Debian, the
> > BSD folks, etc. No idea about Windows and OSX.
> > What I tried to communicate here was, the instructions might be wrong
> > and misleading. You didn't respond to my question, so I'll just leave
> > you with this.
> I guess I don't understand what you're looking for then.
> 
> -- 
> Peter Šurda
> Bitmessage core developer

What I tried to tell you in the previous message: Personally I'm done with this
conversation, and I hope you don't repeat what happened last year and we
can just do useful things with our time instead of arguing. Thanks.
-- 
ng0 -- https://www.inventati.org/patternsinthechaos/