perlpolicy - Various and sundry policies and commitments related to the perl core
This document is the master document which records all written policies about how the Perl 5 Porters collectively develop and maintain the Perl core.
Our community has a long-held belief that backward-compatibility is a virtue, even when the functionality in question is a design flaw.
We would all love to unmake some mistakes we've made over the past decades. Living with every design error we've ever made can lead to painful stagnation. Unwinding our mistakes is very, very difficult. Doing so without actively harming our users is nearly impossible.
Lately, ignoring or actively opposing compatibility with earlier versions of Perl has come into vogue. Sometimes, a change is proposed which wants to usurp syntax which previously had another meaning. Sometimes, a change wants to improve previously-crazy semantics.
Down this road lies madness.
Requiring end-user programmers to change just a few language constructs, even language constructs which no well-educated developer would ever intentionally use is tantamount to saying "you should not upgrade to a new release of Perl unless you have 100% test coverage and can do a full manual audit of your codebase." If we were to have tools capable of reliably upgrading Perl source code from one version of Perl to another, this concern could be significantly mitigated.
We want to ensure that Perl continues to grow and flourish in the coming years and decades, but not at the expense of our user community.
Existing syntax and semantics should only be marked for destruction in very limited circumstances. If a given language feature's continued inclusion in the language will cause significant harm to the language or prevent us from making needed changes to the runtime, then it may be considered for deprecation.
Any language change which breaks backward-compatibility should be able to be enabled or disabled lexically. Unless code at a given scope declares that it wants the new behavior, that new behavior should be disabled. Which backward-incompatible changes are controlled implicitly by a 'use v5.x.y' is a decision which should be made by the pumpking in consultation with the community.
When a backward-incompatible change can't be toggled lexically, the decision to change the language must be considered very, very carefully. If it's possible to move the old syntax or semantics out of the core language and into XS-land, that XS module should be enabled by default unless the user declares that they want a newer revision of Perl.
Historically, we've held ourselves to a far higher standard than backward-compatibility -- bugward-compatibility. Any accident of implementation or unintentional side-effect of running some bit of code has been considered to be a feature of the language to be defended with the same zeal as any other feature or functionality. No matter how frustrating these unintentional features may be to us as we continue to improve Perl, these unintentional features often deserve our protection. It is very important that existing software written in Perl continue to work correctly. If end-user developers have adopted a bug as a feature, we need to treat it as such.
New syntax and semantics which don't break existing language constructs and syntax have a much lower bar. They merely need to prove themselves to be useful, elegant, well designed and well tested.
To make sure we're talking about the same thing when we discuss the removal of features or functionality from the Perl core, we have specific definitions for a few words and phrases.
If something in the Perl core is marked as experimental, we may change its behaviour, deprecate or remove it without notice. While we'll always do our best to smooth the transition path for users of experimental features, you should contact the perl5-porters mailinglist if you find an experimental feature useful and want to help shape its future.
If something in the Perl core is marked as deprecated, we may remove it from thecore in the next stable release series, though we may not. As of Perl 5.12, deprecated features and modules warn the user as they're used. If you use a deprecated feature and believe that its removal from the Perl core would be a mistake, please contact the perl5-porters mailinglist and plead your case. We don't deprecate things without a good reason, but sometimes there's a counterargument we haven't considered. Historically, we did not distinguish between "deprecated" and "discouraged" features.
From time to time, we may mark language constructs and features which we consider to have been mistakes as discouraged. Discouraged features aren't candidates for removal in the next major release series, but we may later deprecate them if they're found to stand in the way of a significant improvement to the core.
Once a feature, construct or module has been marked as deprecated for a stable release cycle, we may remove it from the core. Unsurprisingly, we say we've removed these things.
New releases of maint should contain as few changes as possible. If there is any question about whether a given patch might merit inclusion in a maint release, then it almost certainly should not be included.
Portability fixes, such as changes to Configure and the files in hints/ are acceptable. Ports of Perl to a new platform, architecture or OS release that involve changes to the implementation are NOT acceptable.
Documentation updates are acceptable.
Patches that add new warnings or errors or deprecate features are not acceptable.
Patches that fix crashing bugs that do not otherwise change Perl's functionality or negatively impact performance are acceptable.
Patches that fix CVEs or security issues are acceptable, but should be run through the perl5-security-report@perl.org mailing list rather than applied directly.
Updates to dual-life modules should consist of minimal patches to fix crashing or security issues (as above).
New versions of dual-life modules should NOT be imported into maint. Those belong in the next stable series.
Patches that add or remove features are not acceptable.
Patches that break binary compatibility are not acceptable. (Please talk to a pumpking.)
Historically, only the pumpking cherry-picked changes from bleadperl into maintperl. This has...scaling problems. At the same time, maintenance branches of stable versions of Perl need to be treated with great care. To that end, we're going to try out a new process for maint-5.12.
Any committer may cherry-pick any commit from blead to maint-5.12 if they send mail to perl5-porters announcing their intent to cherry-pick a specific commit along with a rationale for doing so and at least two other committers respond to the list giving their assent. (This policy applies to current and former pumpkings, as well as other committers.)
What follows is a statement about artistic control, defined as the ability of authors of packages to guide the future of their code and maintain control over their work. It is a recognition that authors should have control over their work, and that it is a responsibility of the rest of the Perl community to ensure that they retain this control. It is an attempt to document the standards to which we, as Perl developers, intend to hold ourselves. It is an attempt to write down rough guidelines about the respect we owe each other as Perl developers.
This statement is not a legal contract. This statement is not a legal document in any way, shape, or form. Perl is distributed under the GNU Public License and under the Artistic License; those are the precise legal terms. This statement isn't about the law or licenses. It's about community, mutual respect, trust, and good-faith cooperation.
We recognize that the Perl core, defined as the software distributed with the heart of Perl itself, is a joint project on the part of all of us. From time to time, a script, module, or set of modules (hereafter referred to simply as a "module") will prove so widely useful and/or so integral to the correct functioning of Perl itself that it should be distributed with Perl core. This should never be done without the author's explicit consent, and a clear recognition on all parts that this means the module is being distributed under the same terms as Perl itself. A module author should realize that inclusion of a module into the Perl core will necessarily mean some loss of control over it, since changes may occasionally have to be made on short notice or for consistency with the rest of Perl.
Once a module has been included in the Perl core, however, everyone involved in maintaining Perl should be aware that the module is still the property of the original author unless the original author explicitly gives up their ownership of it. In particular:
The version of the module in the core should still be considered the work of the original author. All patches, bug reports, and so forth should be fed back to them. Their development directions should be respected whenever possible.
Patches may be applied by the pumpkin holder without the explicit cooperation of the module author if and only if they are very minor, time-critical in some fashion (such as urgent security fixes), or if the module author cannot be reached. Those patches must still be given back to the author when possible, and if the author decides on an alternate fix in their version, that fix should be strongly preferred unless there is a serious problem with it. Any changes not endorsed by the author should be marked as such, and the contributor of the change acknowledged.
The version of the module distributed with Perl should, whenever possible, be the latest version of the module as distributed by the author (the latest non-beta version in the case of public Perl releases), although the pumpkin holder may hold off on upgrading the version of the module distributed with Perl to the latest version until the latest version has had sufficient testing.
In other words, the author of a module should be considered to have final say on modifications to their module whenever possible (bearing in mind that it's expected that everyone involved will work together and arrive at reasonable compromises when there are disagreements).
As a last resort, however:
If the author's vision of the future of their module is sufficiently different from the vision of the pumpkin holder and perl5-porters as a whole so as to cause serious problems for Perl, the pumpkin holder may choose to formally fork the version of the module in the core from the one maintained by the author. This should not be done lightly and should always if at all possible be done only after direct input from Larry. If this is done, it must then be made explicit in the module as distributed with Perl core that it is a forked version and that while it is based on the original author's work, it is no longer maintained by them. This must be noted in both the documentation and in the comments in the source of the module.
Again, this should be a last resort only. Ideally, this should never happen, and every possible effort at cooperation and compromise should be made before doing this. If it does prove necessary to fork a module for the overall health of Perl, proper credit must be given to the original author in perpetuity and the decision should be constantly re-evaluated to see if a remerging of the two branches is possible down the road.
In all dealings with contributed modules, everyone maintaining Perl should keep in mind that the code belongs to the original author, that they may not be on perl5-porters at any given time, and that a patch is not official unless it has been integrated into the author's copy of the module. To aid with this, and with points #1, #2, and #3 above, contact information for the authors of all contributed modules should be kept with the Perl distribution.
Finally, the Perl community as a whole recognizes that respect for ownership of code, respect for artistic control, proper credit, and active effort to prevent unintentional code skew or communication gaps is vital to the health of the community and Perl itself. Members of a community should not normally have to resort to rules and laws to deal with each other, and this document, although it contains rules so as to be clear, is about an attitude and general approach. The first step in any dispute should be open communication, respect for opposing views, and an attempt at a compromise. In nearly every circumstance nothing more will be necessary, and certainly no more drastic measure should be used until every avenue of communication and discussion has failed.
Social Contract about Contributed Modules originally by Russ Allbery <rra@stanford.edu> and the perl5-porters.