Re: About tidying up Kwalitee metrics
* On Sun, Jun 29 2008, chromatic wrote: However, does making CPAN a better place require publishing a Hall of Shame on perl.org? http://cpants.perl.org/highscores/hall_of_shame Good point. The same could be said for CPAN Ratings also. Why should my module have 1 star next to it because any goof with a web browser can write a review? Why is the opinion of someone with no ties to the community considered relevant enough to show in the search.cpan search results? (The same goes for positive ratings. I've seen a lot of high ratings of modules that are rated highly for no good reason, or rated that way by its own author.) I personally don't care and generally ignore the ratings, but it's the same thing as Kwalitee, except not even objective. Regards, Jonathan Rockway -- print just = another = perl = hacker = if $,=$
Re: About tidying up Kwalitee metrics
On Jun 30, 2008, at 1:08 AM, Jonathan Rockway wrote: Why is the opinion of someone with no ties to the community considered relevant enough to show in the search.cpan search results? Why do you think the opinion of someone with ties to the community (however THAT is defined) is more relevant than someone who doesn't? Our little echo chamber is not some hallowed hall that indicates programming wisdom. xoa -- Andy Lester = [EMAIL PROTECTED] = www.petdance.com = AIM:petdance
Re: About tidying up Kwalitee metrics
On Sunday 29 June 2008 23:08:50 Jonathan Rockway wrote: * On Sun, Jun 29 2008, chromatic wrote: However, does making CPAN a better place require publishing a Hall of Shame on perl.org? http://cpants.perl.org/highscores/hall_of_shame Good point. The same could be said for CPAN Ratings also. Why should my module have 1 star next to it because any goof with a web browser can write a review? Why is the opinion of someone with no ties to the community considered relevant enough to show in the search.cpan search results? (The same goes for positive ratings. I've seen a lot of high ratings of modules that are rated highly for no good reason, or rated that way by its own author.) I personally don't care and generally ignore the ratings, but it's the same thing as Kwalitee, except not even objective. There are important differences. CPAN Ratings are much more obviously subjective. No one (so far) has ranked all 16,000 or however many CPAN distributions against each other in a canonical list. Ratings have individual names attached to them. They're not just perl.org says that these X distributions from these Y authors are particularly shameful. (Note that the Hall of Shame doesn't include the Kwalitee is not Quality dodge. Then again, neither does the Hall of Triumph.) Ratings have text that people can read and analyze on their own, if they want. None of these mean that potential users *will* use all of their tools, but the differences seem important to me. -- c
Re: CPAN Ratings and the problem of choice (was Re: About tidying up Kwalitee metrics)
Paul Fenwick said: Jonathan Rockway wrote: The same could be said for CPAN Ratings also. Why should my module have 1 star next to it because any goof with a web browser can write a review? Why is the opinion of someone with no ties to the community considered relevant enough to show in the search.cpan search results? I'm a big supporter of CPAN Ratings, because I view them as solving one of the biggest problems facing the CPAN today. Choice overload. CPAN is suffering from its own success. One of the most common questions I get asked is Which CPAN module should I use? There's like 300 that cover my problem. The worst thing is, faced with too many choices, typical humans are more likely to choose *none* of them, compared with if they were only offered one or two[1]. Thank you for making this point. I've had this problem too, many times, and I'd love to see something that helps me manage it. Let's assume I'm in hurry to buy a present to someone I don't know (or any other situation where I'm forced to make a low-info, low-context decision.) I have to make the best out of the situation with the information that I have. Sometimes the only solution is just to ask the clerk what toy would you give as a birthday present to a 5 year-old friend of your nephew?. The clerk would at least be able to give _some_ useful info, like this is popular amongst the pre-schoolers or this toy got a prize for being the most educational in 2007 or we are getting lots of these toys in return, so don't buy it until the problems are fixed upstream... The criteria for choosing software are of course a bit different. I'd argue the major one is that WE can also choose to improve the software we select (at least when it comes to OSS.) So when we're discussing Kwalitee metrics or the CPANTS game, we're in fact discussing new datapoints for people to use when they choose. We make information available. We're communicating. But as with all other kinds of communication, we have both transmitters of information (the CPANTS website, metrics, explanations, reviews etc.) and a receiver (the individual end users, the distro authors), and as with all other kinds of communication, there's always a danger for the recipient to interpret the info wrong. There's a tradition in the marketing and sales professions that if a message doesn't land well, then one should assume something is wrong with the message, and not the recipient. This may be well and true in most cases, but it doesn't take much to imagine situations where this assumption is wrong - or at least not precise enough. But for our purposes I think this tradition would apply well. If people are actually annoyed about getting in the hall of shame, we shouldn't remove the hall, but instead give them useful info on how to get out of it. If authors add useless workarounds just to get on the top of the CPANTS game, we shouldn't remove the game, but instead find ways to make this tactic useless. It's extremely telling when one of the most popular parts of Perl Training Australia's courses is showing students the Phalanx 100 as a short-list. Even though the list is quite some years old, there's almost palpable relief when the students realise they can just pick XML::Parser from the Phalanx top 10, rather than having to examine the multitude of choices on the CPAN. So, why do ratings make a difference here? Well, ratings provide at least a partial way for the community to solve the choice overload problem. If a search reveals a 4.5 star module with eight reviews, one doesn't feel compelled to look at the other options; the choice becomes clear. Let's look at one assumption I think we're making... Who are actually the information recipients in this matter? Here's my take on it: * End users of CPAN modules * CPAN module authors x People who are in a learning mode x People who are in a getting things done mode So, who should we tailor the messages for? Here's how I would rank the message recipients: 1. End users of CPAN modules who are in a getting things done mode (help users choose, because this makes CPAN into Perl's killer app) 2. CPAN module authors who are in a learning mode (help authors make better modules, because we want less than 90% crap) 3. End users of CPAN modules who are in a learning mode (help users become authors, because this is how the community grows) 4. CPAN module authors who are in a getting things done mode. (help authors work efficiently/without annoyances) If we can agree on this, I think it'll be a lot easier to decide of ways and means to move CPAN forward, and even make some good decisions. - Salve -- #!/usr/bin/perl sub AUTOLOAD{$AUTOLOAD=~/.*::(\d+)/;seek(DATA,$1,0);print# Salve Joshua Nilsen getc DATA}$='};{';@_=unpack(C*,unpack(u*,':4@,$'.# [EMAIL PROTECTED] '2!--5-(50P%$PL,!0X354UC-PP%/0\`'.\n));eval {'@_'}; __END__ is near! :)
Re: CPAN Ratings and the problem of choice (was Re: About tidying up Kwalitee metrics)
So, why do ratings make a difference here? Well, ratings provide at least a partial way for the community to solve the choice overload problem. If a search reveals a 4.5 star module with eight reviews, one doesn't feel compelled to look at the other options; the choice becomes clear. I question the usefulness of the ratings because they are almost completely unused. The module mentioned in this thread, XML::Parser, has 6 reviews (2 of which are basically bug reports, and one tells you to not use it for any new code). One of the oldest and most important modules ever, DBI, has a mere 29. That's 29 reviews in 8 years - pathetic. It should have hundreds of ratings. The important question to ask is, (assuming the ratings are something worth keeping), why are people not rating modules, and how can we encourage people to do so? -- Greg Sabino Mullane [EMAIL PROTECTED] End Point Corporation signature.asc Description: PGP signature
Re: About tidying up Kwalitee metrics
--- On Sat, 28/6/08, Aristotle Pagaltzis [EMAIL PROTECTED] wrote: I think the game is actually an excellent idea. The problem is with the metrics. Here are some metrics that are inarguably good: • has_buildtool • extracts_nicely • metayml_conforms_to_known_spec One problem with this is when you get dinged for an unknown key. This means you can't extend your meta YAML file. It's a hash disguised as YAML. There shouldn't be a problem with adding to it, only subtracting from it. On a side note, I still don't understand why I sometimes get dinged for CPANTs errors. Here's one for HOP-Lexer (http://cpants.perl.org/dist/errors/HOP-Lexer): STDERR: Invalid row in Debian file: libhtml-wikiconverter-moinmoin-perl, HTM STDOUT: I have no idea what this is and I have no way of correcting it yet I am getting dinged for it. I see that I can send an email to [EMAIL PROTECTED], but why? I don't understand why CPANTs bugs are counted against me. Cheers, Ovid
Re: About tidying up Kwalitee metrics
Hi! On Sun, Jun 29, 2008 at 01:54:19AM -0700, Ovid wrote: On a side note, I still don't understand why I sometimes get dinged for CPANTs errors. Here's one for HOP-Lexer (http://cpants.perl.org/dist/errors/HOP-Lexer): STDERR: Invalid row in Debian file: libhtml-wikiconverter-moinmoin-perl, HTM STDOUT: I have no idea what this is and I have no way of correcting it yet I am getting dinged for it. I see that I can send an email to [EMAIL PROTECTED], but why? I don't understand why CPANTs bugs are counted against me. As Gabor already suggested, most of the texts on cpants.perl.org should be overhauled and extendend. For example: http://cpants.perl.org/kwalitee.html#no_cpants_errors no_cpants_errors Shortcoming: Some errors occured during CPANTS testing. They might be caused by bugs in CPANTS or some strange features of this distribution. See 'cpants' in the dist error view for more info. Remedy: Please report the error(s) to [EMAIL PROTECTED] 'Shortcoming' should be extended to say: The goal of deducting a kwalitee point for 'no_cpants_errors' is to get authors to report CPANTS bugs. As you might guess, testing 10.000+ different dists is hard. There are lot of special cases. It's impossible to figure out all those special cases in advance. 'no_cpants_errors' is a way to outsource the discovery of special cases to module authors. or something like that... -- #!/usr/bin/perl http://domm.plix.at for(ref bless{},just'another'perl'hacker){s-:+-$-gprint$_.$/}
Re: About tidying up Kwalitee metrics
* Ovid [EMAIL PROTECTED] [2008-06-29 10:55]: --- On Sat, 28/6/08, Aristotle Pagaltzis [EMAIL PROTECTED] wrote: I think the game is actually an excellent idea. The problem is with the metrics. Here are some metrics that are inarguably good: • has_buildtool • extracts_nicely • metayml_conforms_to_known_spec One problem with this is when you get dinged for an unknown key. This means you can't extend your meta YAML file. It's a hash disguised as YAML. There shouldn't be a problem with adding to it, only subtracting from it. On a side note, I still don't understand why I sometimes get dinged for CPANTs errors. Yes, but that doesn’t detract from my point. If those metrics are faulty, they should and *can* be fixed – and either way they measure good form directly, as good metrics should. The problems with them don’t fall in the same category as looking for arbitrarily chosen proxies for unmeasurable aspects of good form (or even style). Regards, -- Aristotle Pagaltzis // http://plasmasturm.org/
Re: About tidying up Kwalitee metrics
In article [EMAIL PROTECTED], Thomas Klausner [EMAIL PROTECTED] wrote: The goal of deducting a kwalitee point for 'no_cpants_errors' is to get authors to report CPANTS bugs. Why do you need authors to report those? After a run, you have a list of all of the errors already.
Re: About tidying up Kwalitee metrics
On Sunday 29 June 2008 02:28:54 Thomas Klausner wrote: For example: http://cpants.perl.org/kwalitee.html#no_cpants_errors no_cpants_errors Shortcoming: Some errors occured during CPANTS testing. They might be caused by bugs in CPANTS or some strange features of this distribution. See 'cpants' in the dist error view for more info. Remedy: Please report the error(s) to [EMAIL PROTECTED] 'Shortcoming' should be extended to say: The goal of deducting a kwalitee point for 'no_cpants_errors' is to get authors to report CPANTS bugs. As you might guess, testing 10.000+ different dists is hard. There are lot of special cases. It's impossible to figure out all those special cases in advance. 'no_cpants_errors' is a way to outsource the discovery of special cases to module authors. or something like that... I thought the goal of Kwalitee was to identify good free software, not to humiliate thousands of other authors of free software for not anticipating and working around your bugs. I didn't ask you to scan my distributions, and it's kind of a problem for me that you're willing to write publicly that their Kwalitee would be higher if I reported bugs in code I didn't write, don't use, and don't believe in -- especially if you're going to claim that Kwalitee metrics are useful in deciding whether to use my distributions. (If you don't claim that, then replace my objection with Okay, so what's the point again?) Want to fix CPANTS and Kwalitee? It's simple: * get rid of the scoreboard * dump the harmful metrics (POD checking, etc) * separate all of the informational metrics from the genuinely useful metrics * report to authors when their uploads fail the useful metrics -- c
Re: About tidying up Kwalitee metrics
On Sun, Jun 29, 2008 at 4:49 PM, chromatic [EMAIL PROTECTED] wrote: On Sunday 29 June 2008 02:28:54 Thomas Klausner wrote: For example: http://cpants.perl.org/kwalitee.html#no_cpants_errors no_cpants_errors Shortcoming: Some errors occured during CPANTS testing. They might be caused by bugs in CPANTS or some strange features of this distribution. See 'cpants' in the dist error view for more info. Remedy: Please report the error(s) to [EMAIL PROTECTED] 'Shortcoming' should be extended to say: The goal of deducting a kwalitee point for 'no_cpants_errors' is to get authors to report CPANTS bugs. As you might guess, testing 10.000+ different dists is hard. There are lot of special cases. It's impossible to figure out all those special cases in advance. 'no_cpants_errors' is a way to outsource the discovery of special cases to module authors. or something like that... I thought the goal of Kwalitee was to identify good free software, not to humiliate thousands of other authors of free software for not anticipating and working around your bugs. I also think the no_cpants_errors has no place in the core metrics nor actually any metric. It should be only seen by the CPANTS authors ... but chromatic, while I have not added that specific metric your tone is offending and humiliating me and maybe also Thomas and possibly others who invest time to try to make CPAN a better place. Gabor
Re: About tidying up Kwalitee metrics
On Sunday 29 June 2008 11:02:17 Gabor Szabo wrote: On Sun, Jun 29, 2008 at 4:49 PM, chromatic [EMAIL PROTECTED] wrote: I thought the goal of Kwalitee was to identify good free software, not to humiliate thousands of other authors of free software for not anticipating and working around your bugs. I also think the no_cpants_errors has no place in the core metrics nor actually any metric. It should be only seen by the CPANTS authors ... but chromatic, while I have not added that specific metric your tone is offending and humiliating me and maybe also Thomas and possibly others who invest time to try to make CPAN a better place. I certainly don't mean to humiliate anyone. Please accept my apologies. However, does making CPAN a better place require publishing a Hall of Shame on perl.org? http://cpants.perl.org/highscores/hall_of_shame I think what I want from CPANTS is conceptually simple: * tell me (and my potential users) if a recent upload is well-behaved (all but three of the core metrics achieve this) * provide optional information as information alone (packaged by various OS distributions, used by other CPAN distributions) * drop the game, with winners and losers and (especially) scores -- c
Re: About tidying up Kwalitee metrics
On Thu, Jun 26, 2008 at 2:23 AM, Hilary Holz [EMAIL PROTECTED] wrote: On 6/25/08 10:24 AM, chromatic [EMAIL PROTECTED] wrote: On Wednesday 25 June 2008 03:15:59 Thomas Klausner wrote: One comment regarding 'each devel sets his/her own kwalitee metrics': This could be quite easy for the various views etc. But I'm not sure how to calculate a game score then. Do we end up with lots of different games? But then, it's only the game (which still motivates a few people..) Removing the game score completely would fix a lot of what I consider wrong with CPANTS. -- c second! It seems that the game theme is after all turned into fierce competition or lack of interest depending on ... I don't know on what, but neither is good for CPAN. In some cases - me included - people fix the symptom to get the metric point while the underlying code does not really change. So the indicator stops being an indicator. I don't know how to fix that. Maybe the suggestions above and elsewhere to get rid of the game theme and the top N bottom N authors would help. Maybe what we need to do is 1) remove the game 2) fix the current metrics (e.g. license is not correct now) 3) Add detailed explanations for each metric, or maybe to create a page on the TPF Perl 5 wiki for each metric where it would be easier to provide pro and contra explanations for each metric. 4) add more metrics (including those that collect data from external sources) 5) categorize the metrics as suggested by Salve 6) get the search engines to start to use some of the metrics in their search results. Not necessarily in that order Gabor
Re: About tidying up Kwalitee metrics
On Saturday 28 June 2008 08:54:34 Gabor Szabo wrote: It seems that the game theme is after all turned into fierce competition or lack of interest depending on ... I don't know on what, but neither is good for CPAN. In some cases - me included - people fix the symptom to get the metric point while the underlying code does not really change. So the indicator stops being an indicator. Exactly -- and in other cases, the metric point is actively harmful to the CPAN. Maybe what we need to do is 1) remove the game 2) fix the current metrics (e.g. license is not correct now) 3) Add detailed explanations for each metric, or maybe to create a page on the TPF Perl 5 wiki for each metric where it would be easier to provide pro and contra explanations for each metric. 4) add more metrics (including those that collect data from external sources) 5) categorize the metrics as suggested by Salve 6) get the search engines to start to use some of the metrics in their search results. Not necessarily in that order Full support from me on these. -- c
Re: About tidying up Kwalitee metrics
Hi! On Tue, Jun 24, 2008 at 10:10:07AM +0200, Salve J Nilsen wrote: I propose to split the current main and optional kwalitee scales into topical ones, so we can allow for richer set of metrics while allowing everyone that care mostly about certain types of metric access to untainted versions. ... Thoughts? I've been very quite lately regarding CPANTS, mostly because I currently have more interesting things to do (at the moment I'm in the lucky situation that my day job is more fun than my non-paid open source activities). This does not mean that I want to give up maintaing CPANTS. I like most of the feedback given here (and on use.perl) in the last months, and would love to turn some of the suggestions into code. I would of course love it even more, if you [all of you, not Slave] would turn some of the suggestions into code... Anyway, next week my kids are on holidays with my father, which will buy me some extra time. Some of this time will go into CPANTS. I have a talk on CPANTS scheduled for YAPC::Europe. I would love to turn this into a how to contribute thing, followed by a hacking-session. One comment regarding 'each devel sets his/her own kwalitee metrics': This could be quite easy for the various views etc. But I'm not sure how to calculate a game score then. Do we end up with lots of different games? But then, it's only the game (which still motivates a few people..) Oh, and of course even yet only 'core' metrics are used to calculate the game score. Anyway, even if I do not reply on all comments, I'm collecting the feedback and will comment/implement parts of it. When I have tuits. -- #!/usr/bin/perl http://domm.plix.at for(ref bless{},just'another'perl'hacker){s-:+-$-gprint$_.$/}
Re: About tidying up Kwalitee metrics
On Tue, Jun 24, 2008 at 10:10:07AM +0200, Salve J Nilsen wrote: Hello, folks I propose to split the current main and optional kwalitee scales into topical ones, so we can allow for richer set of metrics while allowing everyone that care mostly about certain types of metric access to untainted versions. Let's remove the optional type, and instead create the following metrics where we can place the existing tests: Disto Kwalitee (most of the original test should go here) Security Kwalitee (checks for taint-mode or other security-related issues go here) Community Support Kwalitee (checks for supplied mailing list address, bugtracker, archives, etc. go here) Community Trust Kwalitee (analysis of external acceptance of the module, including Debian use go here) Thoughts? Certainly, I would like the metrics to be split into those I can control by what I upload to PAUSE, and those that I can't fix however much I upload. Which I think most obviously is those that you group here as Community Trust Kwalitee. The previous 2 seem good, as they are likely to be categories that some people have legitimate disagreements with. ie I've not been paying close attention to CPANTS, but if I did, I suspect that it would annoy me that it expects me to have a POD coverage test, and that in turn to make it pass I could well spend more time bodging that than actually writing documentation. Which, I agree with chromatic, would be stupid, and not something that I'd like to see promoted. (Is You have POD and it's well formed is something that is already tested?) Nicholas Clark
Re: About tidying up Kwalitee metrics
G'day Thomas, Thomas Klausner wrote: I've been very quite lately regarding CPANTS, mostly because I currently have more interesting things to do (at the moment I'm in the lucky situation that my day job is more fun than my non-paid open source activities). This does not mean that I want to give up maintaing CPANTS. Congratulations! ;) would of course love it even more, if you [all of you, not Slave] would turn some of the suggestions into code... In other words, put our code where our mouths are. ;) For those of you who want hacking the CPANTS game to be a game in itself, it now earns you ohloh kudos too: http://www.ohloh.net/projects/cpants This could be quite easy for the various views etc. But I'm not sure how to calculate a game score then. Do we end up with lots of different games? But then, it's only the game (which still motivates a few people..) I've tossed out a few of these suggestions, so I guess I better start coming up with answers. I'll start with what I think are the least controversial things, and get into more risky territory as I go. == Honours == One of the proposals was that some of the optional metrics like packaged by Debian become honours. These are things which are (more-or-less) out of the author's control, but which we already have (disabled) tests for, and which are useful indicators of quality. I suggest that completed honours are shown automatically for any distribution that has them. Honours that a distribution doesn't have just don't get shown. == Optional Metrics == I've also proposed that things that the author does have control over, but which they don't consider relevant to their distribution(s), can be switched off. For the optional metrics these allegedly don't contribute to the game score[1], and so the ability to disable them *should* be a non-issue; you don't gain or lose game rankings by having them or not. Optional metrics that an author doesn't want are simply not shown. == Kwalitee Scores == Getting a little more controversial here, this means splitting the Kwalitee score into two. Rather than showing an aggregate Kwalitee score for each distribution, we'd show the Core Kwalitee (for non-optional metric) and the Bonus Kwalitee. If you're a gamer, you'll be turning on bonus kwalitee metrics and trying to complete them to obtain the fabled Amulet of CPANTS. If you're not a gamer, you'll turn them off, and that's that. Whether that means we have *two* scoreboards I'll leave as an open question, but I'd be willing to bet the general consensus is that we should. Cheerio, Paul [1] However it appears that http://cpants.perl.org/highscores/hall_of_fame has scores going up to 128, which I understand means it *does* includes optional metrics. http://cpants.perl.org/dist/overview/IPC-System-Simple has a score of 124, and fails only one optional metric (which it would have passed if I had remembered 'make manifest'). -- Paul Fenwick [EMAIL PROTECTED] | http://perltraining.com.au/ Director of Training | Ph: +61 3 9354 6001 Perl Training Australia| Fax: +61 3 9354 2681
Re: About tidying up Kwalitee metrics
On Wednesday 25 June 2008 03:15:59 Thomas Klausner wrote: One comment regarding 'each devel sets his/her own kwalitee metrics': This could be quite easy for the various views etc. But I'm not sure how to calculate a game score then. Do we end up with lots of different games? But then, it's only the game (which still motivates a few people..) Removing the game score completely would fix a lot of what I consider wrong with CPANTS. -- c