A discussion of DBIx-Class governance and future development

Acting in my capacity as an administrator for PAUSE, I've been mediating a dispute over the future disposition of primary permissions for the DBIx::Class namespace on CPAN. I recently posted a message to the mailing list for DBIx::Class titled "IMPORTANT: A discussion of DBIC governance and future development".

I am reprinting it in full below in the hope that doing so will help this message reach DBIx::Class users who are not on the mailing list. I encourage such users to read the message and join the mailing list to participated in the conversation and express their interests.

Subject: IMPORTANT: A discussion of DBIC governance and future development

Hello, DBIC community.

I apologize in advance for the length of this email, but I urge everyone that uses DBIC to read it fully as it relates to the future of this important module.

For those who don't know me, I'm DAGOLDEN on CPAN and I've joined this list in my capacity as a PAUSE [1] administrator.

For those on the list who aren't familiar with CPAN administration, PAUSE is the service that authors use to upload modules to CPAN. Among other functions, it generates the index that maps modules names to downloadable tarballs – e.g. "DBIx::Class" to "RIBASUSHI/DBIx-Class-0.082840.tar.gz" on a CPAN mirror.

PAUSE also maintains a permissions model [2] for each module namespace with two levels: "primary maintainer" (also called "first come") and "co-maintainer" (aka "co-maint"). Primary maintainers can grant and revoke co-maint permissions. Both levels can upload tarballs to PAUSE, triggering an update to the index.

Over the past several weeks, I've been the PAUSE administrator selected to mediate a dispute over future disposition of primary permissions for the DBIx::Class namespace.

The dispute was triggered by Peter Rabbitson's "Traffic pattern changes ahead" [3] email to this list on September 6, in which he said:

I have also firmly selected who will be getting the DBIx::Class
namespace first-come, the transfer of which will also happen
somewhere around the end of September.

Because the identity of the new primary maintainer was neither disclosed nor discussed with Matt Trout (the founder of the DBIC project, current co-maintainer and also PAUSE administrator) or other co-maintainers, several private conversations between ensued between Matt, Peter and others about this declaration.

On September 15, Peter notified PAUSE administrators via the modules@perl.org mailing list of an "Upcoming PAUSE permissions dispute" [4]. Separately, Matt notified PAUSE administrators privately with his own concerns about a possible dispute (his email was later disclosed and I'll link to it later).

On September 21, I privately emailed all DBIC maintainers (CPAN authors ABRAXXA, ARODLAND, FREW, ILMARI, JROBINSON, MSTROUT, and RIBASUSHI) on behalf of PAUSE administrators with our collective view of how this dispute would be best resolved. Peter asked that any discussion be public, so I reposted it to the modules@perl.org mailing list as "Message from PAUSE Admins to DBIx::Class maintainers [resend]" [5]

I urge everyone to read that thread in full as well. For reference, it includes a copy [6] of Matt's previously private email to PAUSE administrators.

Importantly, the thread summarizes PAUSE administrators' position on the dispute, which I repost verbatim here:

  1. Given the importance of DBIC to the broader Perl community (i.e. way "upriver" <http://neilb.org/2015/04/20/river-of-cpan.html>), we’d like to see a more open discussion between existing maintainers about what happens next in terms of DBIC leadership and control of primary permissions.
  2. The best outcome from our perspective would be for a successor to be decided by consensus of existing maintainers.
  3. Given a dispute among maintainers, the only outcome that is absolutely unacceptable to PAUSE admins would be a unilateral permissions transfer decision.
  4. We really hope the DBIC maintainers and/or community can resolve this internally.

In the ensuing discussion, Peter disclosed additional details about his plans for the future of DBIC in the "Future plans" section of this email [7]:

I am still planning to wrap up the remaining pieces, including some
unannounced initiatives to get the project into the best shape possible
to survive its leaderlessness.

I am still planning to remove all co-maint perms and handover the
first-come to a yet-undisclosed person. Given no clear line of
succession, and the incredibly high stakes wrt compatibility, the only
responsible thing to do is to select a single spot of responsibility and
provide all possible support and infrastructure for a proper
project-freeze.

In another email [8], Peter suggested raising these issues explicitly on the DBIC mailing list:

As suggested in an earlier email: the PAUSE admins (as the only
legitimate concerned party at this point) would likely benefit having
this question asked in a wider forum ( the DBIC mailing list and/or
other channels ). Essentially someone has to trigger a "vote of no
confidence", otherwise this entire exchange is just a time consuming farce.

On behalf of the PAUSE administrators, we would therefore like to invite Peter to describe in more detail his plans for a "project freeze" and the role he envisions for a successor maintainer. We invite Matt, other co-maintainers, and the DBIC community at large to add their thoughts about the specifics of the plan or about the situation in general.

Given public and private discussions to date, we believe the DBIC community should consider questions such as:

  • How should the future governance of the DBIC project be decided?
  • Who should or shouldn't be involved in future governance?
  • Should the project be "frozen" or should development continue?
  • If "frozen", what specifically would a "freeze" entail? Would there be exceptions?
  • If not "frozen", what principles should govern development? (Cathedral vs Bazaar [9] and/or New Jersey Style vs MIT Style [10])

We believe these discussions, if had openly, honestly and constructively, will lead to the best resolution of this dispute for the DBIC community.

Thank you for reading this far, and I look forward to reading the community's views on these matters.

Sincerely,
David Golden, PAUSE Administrator

[1] http://pause.perl.org/
[2] http://perladvent.org/2013/2013-12-08.html
[3] http://lists.scsys.co.uk/pipermail/dbix-class/2016-September/012187.html
[4] http://www.nntp.perl.org/group/perl.modules/2016/09/msg96115.html
[5] http://www.nntp.perl.org/group/perl.modules/2016/09/msg96139.html
[6] http://www.nntp.perl.org/group/perl.modules/2016/10/msg96178.html
[7] http://www.nntp.perl.org/group/perl.modules/2016/10/msg96174.html
[8] http://www.nntp.perl.org/group/perl.modules/2016/10/msg96182.html
[9] https://en.wikipedia.org/wiki/The_Cathedral_and_the_Bazaar
[10] https://en.wikipedia.org/wiki/Worse_is_better

Posted in perl programming | Tagged , , , | Comments closed

Comparison of Class::Tiny and Object::Simple

Yuki Kimoto recently posted about the latest release of Object::Simple, billed as "the simplest class builder". Since I've also written a "simple" OO framework called Class::Tiny, I thought I'd point out similarities and differences.

(I'm not going to address Object::Simple's origins from or differences from Mojo::Base.)

Similarities

Single file, minimal dependencies

Both Object::Simple and Class::Tiny are single-file OO frameworks with no no-core dependencies on recent perls. According to the "sloccount" tool, Object::Simple is 98 lines. Class::Tiny is 135.

Class::Tiny does require some dependencies on older Perls for deep @ISA introspection and global destruction detection.

Accessor generation with lazy defaults

Both frameworks allow you to specify accessors and provide either scalar or code-reference defaults for them. Defaults are evaluated on first use. The underlying generated code is extraordinarily similar and accessor speeds are generally comparable (at least with Class-Tiny-1.05 which has some optimizations to remove scopes).

Read-write accessors

Both offer read-write accessors, which I think is the only sensible choice when providing only a single style.

Differences

Mutator return style

Class::Tiny mutators return the value just set, which is consistent with the values returned by accessors. Object::Simple mutators return the invocant, which allows chaining.

BUILD/BUILDARGS/DEMOLISH

Class::Tiny supports the BUILD/BUILDARGS/DEMOLISH methods just like Moose and Moo do. Object::Simple does not.

Notably, Class::Tiny supports an interoperability convention that allows Moo or Moose classes to inherit from a Class::Tiny class without calling BUILD methods more than once.

Constructor speed

Because Class::Tiny does some extra validation, plus provides BUILD/BUILDARGS support, its constructor is about 3x slower than Object::Simple, which has a two-line constructor.

Extraneous methods in @ISA

Class::Tiny classes inherit from Class::Tiny::Object, which provides only new, BUILDALL and DESTROY methods. Object::Simple classes typically inherit from Object::Simple, which provides import, new, attr, class_attr and dual_attr methods.

Unknown constructor arguments

Class::Tiny ignores unknown attributes in constructor arguments (without error, just like Moose/Moo). Object::Simple will include them in the constructed object.

Subclassing

Class::Tiny relies on users to set inheritance with @ISA or base/super/parent pragmas. Object::Simple additionally offers an import flag "-base" which sets the superclass. If the superclass is not Object::Simple, the superclass is loaded.

Introspection

Class::Tiny provides a mechanism for getting a list of class attributes and default values for attributes. Object::Simple does not.

strict/warnings export

Object::Simple turns on strict and warnings in the caller when the "-base" flag is used. Class::Tiny does not.

Closing thoughts

Object::Simple is, indeed, simple. It's not much more than syntactic sugar for generating accessors with defaults.

That said, I think it's too simple. If you really need minimal overhead and maximum speed just bless a hash reference into a class and directly access the members. If you want minimalism and default values, you can get there with eager defaults like this:

sub new {
    my $class = shift;
    return bless { name => "Jane", data => {}, @_ }, $class;
}

Once you start subclassing, I think you'll want BUILD/DEMOLISH support to properly order construction and teardown and Object::Simple doesn't give it to you.

Even if you don't plan to subclass, might that be something your downstream users might want to do? Providing BUILD/DEMOLISH support makes it easy for downstream users to have well-structured construction and teardown.

Yes, you can create custom constructors, but that defeats the syntactic simplicity of Object::Simple. Plus, if you have custom constructors, you'll need custom destructors and a mechanism for ensuring they get called in order. Very soon, you'll have re-invented the semantics of BUILD/DEMOLISH. So why not start with a framework that already provides that for you?

I think Object::Simple fits a very narrow use-case: people who want lazy defaults, don't want to subclass and are willing to add a dependency to avoid some typing.

For general use, I still think Moo is the best all-around choice unless you know for sure that you need the introspection and meta-class hackery that Moose offers.

If Moo is too "heavy" for me for some project, I'll use Class::Tiny. If Class::Tiny is too "heavy" (?!?), then I'll just roll my own class and avoid the dependency entirely.

Admittedly, I'm biased, but I can't think of a situation where I'd actually use Object::Simple as it stands.

If Object::Simple added BUILD/DEMOLISH support, then it might be a decent alternative – a different flavor of simple class builder for those who like its particular API choices (e.g. mutator chaining). Until then, I think it's too niche to put in my toolbox.

Posted in perl programming | Tagged , , | Comments closed

Stand up and be counted: Annual MongoDB Developer Survey

If you use Perl and MongoDB, I need your help. Every year, we put out a survey to developers to find out what languages they use, what features they need, what problems they have, and so on.

We have very few Perl responses. ☹️

Be an ally! Take the MongoDB Developer Experience Survey.

Camel

Posted in perl programming | Tagged , , , | Comments closed

When RFCs attack: HTTP::Tiny is getting stricter

The problem with standards are that there are too many standards. When RFC-2616 – defining HTTP/1.1 – was updated, the IETF spread the details across six RFCs: RFC-7230 to RFC-7235. Most of the changes appear to be server side, but in some areas, particularly around header formatting, the rules are getting tighter.

I was lucky to have a hackathon day at the DC-Baltimore Perl Workshop, so I implemented some of the stricter rules as well as fixing a number of other bugs (and even adding a few features!) See the HTTP::Tiny Changes file for details.

I still need to review more of the RFCs to see if there are other changes that impact HTTP::Tiny.

Meanwhile, HTTP-Tiny-0.057-TRIAL.tar.gz is on CPAN. Please, please, please install it and use it in your day-to-day work to see if there are any problems.

Here is how you can install it with cpanm or cpan:

$ cpanm --dev HTTP::Tiny
$ cpan DAGOLDEN/HTTP-Tiny-0.057-TRIAL.tar.gz

If you find any problems, please open a ticket on the HTTP::Tiny Github issue tracker.

Thanks!

Posted in perl programming | Tagged , , | Comments closed

At-sign nick completion for Weechat and Slack

I like connecting to chat apps like Slack and Flowdock using an IRC client, as that keeps all my chats in one window, right in my terminal.

I've been frustrated that the IRC gateways for Slack and Flowdock don't highlight people based just on their nick the way a typical IRC client does. For example, my IRC nick is "xdg". In an IRC channel, just saying "xdg: ping" will highlight and beep my terminal. But if I want to highlight "bob" who uses the Slack app, I need to say "@bob: ping".

Usually, when I highlight someone, I can use tab completion to finish the nick. That's really handy when I'm talking to my friend "BinGOs". But if I want to @-highlight him and I type "@Bi<TAB>", completion fails.

Since I use Weechat as my IRC client, I whipped up a plugin called atcomplete, which adds an @-prefixed nick for every nick in the nicklist.

With that, I can do "Bi<TAB>" and get "BinGOs", or "<TAB>" again to get "@BinGOs". And if I do "@Bi<TAB>", I get "@BinGOs" right away.

I've submitted it to the Weechat scripts page and it might eventually get approved, but it's available on github now as weechat-atcomplete if people want to try it out. Feedback and patches welcome!

Posted in hacks | Tagged , , , | Comments closed

© 2009-2016 David Golden All Rights Reserved