User: Password:
|
|
Subscribe / Log in / New account

Results of the Apache OpenOffice 4.0 Logo Survey

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 10, 2013 21:41 UTC (Fri) by rcweir (guest, #48888)
In reply to: Results of the Apache OpenOffice 4.0 Logo Survey by mjw
Parent article: Results of the Apache OpenOffice 4.0 Logo Survey

You'd need to do a lot more corrections. For example, with git Ohloh lists all contributors for LO, but with SVN AOO gets credit only for core committers. All contributor patches go through them.

Ohloh also doesn't track AOO branches, where we do most of the work (Ohloh asked us not to include our branches due to the additional load it would put on them). For example, our entire sidebar feature, the cornerstone of the AOO 4.0 work, showed up in Ohloh as a single commit, when the work was integrated into the trunk. But it was a work of many programmers over several months.

Aside from different treatment of HTML editors (which you noticed) there is also differences in how translation work is accounted for, which inflates LO's numbers in Ohloh.

You would also need to consider size of commits, etc. KLoC differences are also hard to compare, since LO did a huge number of very small changes, things translating comments from German to English. Not that this is a bad thing, but it does frustrate simplistic attempts at comparing metrics.

Function points, anyone? No, LO removed a lot of code as well, so you can't compare that.

Or seriously, just realize and accept that these are two different projects and success is what comes out of the project, not what goes into it. And based on that I'm very pleased by our progress. We'll hit 50 million downloads next week. So in terms of the value we bring users, this is huge.

In commercial spaces, no company would brag about how many employees they had or how hard they tried. Bragging rights go to sales. We're not commercial, of course, but it still makes sense to focus on results.


(Log in to post comments)

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 11, 2013 23:49 UTC (Sat) by shmget (subscriber, #58347) [Link]

"there is also differences in how translation work is accounted for, which inflates LO's numbers in Ohloh."

I call bull:
in the last 12 month the 'translation repo' has had about 100 commits.
that is because the work is done in pottle and integrated in the git repo in 'batch'... but you know that very well, since IBM also has setup a pottle server... so you know the workflow.

in comparison you are credited of 910 commits or about 15% of all the commit done (#1 committer) in the last 12 months, none of them touching the code.

"since LO did a huge number of very small changes, things translating comments from German to English."
Bull again:
a quick grep from the log find about 130 commit in lo that are 'translate german' related in the last 12 months... or 0.65%, even if that simplistic grep approach missed half of them... that would still not account for any significant share of the 20K+ commits.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 1:20 UTC (Sun) by rcweir (guest, #48888) [Link]

I think you are mixing up commits with committers. The root of this thread was a claim about committer counts. If you search for "long tail" you'll see my assertion that the long tail doesn't actually account for many commits at all. What you report supports that assertion. Thank you.

The difference between LO and AOO in the case of translators is that LO, by using git, individually lists many of their translators in the logs. (Look at the l10n and dictionaries repos as well). AOO's translators don't touch Subversion. So when you do naive comparisons using log-based reports you count LO translators but not AOO ones.

You'll run into dozens of traps like that when you try to do a naive comparison of LO and AOO, of the type that LO proponents like to propagate. The tool sets and the processes differ enough that such comparisons are very messy. (And looking at the LO clone of AOO's tree in git doesn't solve anything, since that still leaves the process differences, including who checks in code.)

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 2:49 UTC (Sun) by shmget (subscriber, #58347) [Link]

"Look at the l10n and dictionaries repos as well"
First the l10n git repo is obsolete, has been for quite a while
this is now called translations.git
and as I said: "in the last 12 month the 'translation repo' has had about 100 commits."
and since the beginning of the year: 24.

dictionaries which contain... languages dictionaries and thesaurus, has had 34 commits since the beginning of the year.
out of 8200+ commit on master... so both of these represent 0.7% of the commits... not the 'bias' you claim.
btw: the 'translate german comment' stuff represent less than 20 commits this year... all together the 3 things you claim distort the number represent at most 1% of the commits...

"AOO's translators don't touch Subversion."
LO translator do not touch git either...
as I said:
"that is because the work is done in pottle and integrated in the git repo in 'batch'... but you know that very well, since IBM also has setup a pottle server... so you know the workflow."

"You'll run into dozens of traps like that when you try to do a naive comparison of LO and AOO"
no not really... the only things that makes thing harder is _your_ willful attempt to obfuscate things, when, a bit less than a year ago, you decided to _add_ your wiki and website to ohloh as 'source'.
Ironically a few month after having written a blog about how a project can 'boost' their numbers... I guess that is called dog-fooding.

"And looking at the LO clone of AOO's tree in git doesn't solve anything, since that still leaves the process differences, including who checks in code"
yes that does not favor 'over-the-fence' developing...
but the theory that there is a vast variety of author, they are just not committer... well is just a myth. either that or you do not credit the author at all....

Since you choose to use 'since the beginning of the year...

Since the beginning of the year there has been 14 commit in svn that reference
'Patch by' credit (which you said 3 days ago was the way to find these, and I quote: "It might be worth also
looking at names in the SVN commit memo for "Patch by:"

3 of them are credited to people that _are_ committers.
5 of them are credited to the same person, (a paid dev btw)
1 was a helcontent patch
1 was a re-license of an embedded extension
(I mentioned the later 2 because of the claim you made that LO has a lot of small meat-less patches)
1 of them was a dig of an old patch from 2009 out of bugzilla

so all in all a tail length of 6 non-committer authors. that would be missed by looking at git... with a distribution of 5-1-1-1-1-1.

But you already knew all that, since you have discussed it on your dev ML recently...
So as usual you are just hoping that no-one will notice, and purposefully trying to mislead less up-to-date public.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 13:03 UTC (Sun) by rcweir (guest, #48888) [Link]

You are still mixing up commits and committers. Remember, when LibreOffice claims the number of developers, they measure it from the beginning of the project, cumulatively. So even with a trivial % of work (which proves my point about the not-existent long tail)it biases their developer count.

You seem to take offense that we track website content and design changes in Subversion. That's useful for our internal tracking, and we use tools like Ohloh to make that tracking easier. But we have never put that info into a blog post or press release or any other public communication. LibreOffice, on the other hand, frequently promotes data they know is meaningless or deceptive. If you talk to their developers on this, one-on-one, they dismiss it as "the marketing guys doing their thing".

Some examples: LibreOffice claimed, when promoting their community size, the number of wiki accounts they had, even though most of them were never used or just spam accounts:

http://www.robweir.com/blog/2012/11/libreoffices-dubious-...

And similar misrepresentations that inflate their developer counts:

http://www.robweir.com/blog/2012/11/libreoffices-dubious-...

I see how you are trying to accuse AOO of something by tracking website changes, but that argument falls flat because we never promote that number. Only LO supporters promote that number, when trying to draw attention away from their own deceptions.

Finally, but not understanding the process and our use of Subversion you will continue to make false representation of AOO activity. I think the core LO marketing team is smarter in that regard. They avoid doing actual analysis and merely quite Ohloh, which gives them a plausible excuse if their claims are wrong.

Consider this. When LO quotes commits and developers they include the history of whatever code the import into the project from related projects. So when they merge in old branches from OpenOffice.org they count the Oracle developers who worked on that code before LibreOffice even existed. And they count the AOO developers when they merge in their work. We can argue about whether this makes their claimed community statistics more or less accurate, but I hope we agree that an apples-to-apples comparison must credit the merge of 3rd party code similarly.

With AOO 4.0 we have a huge new feature, the sidebar UI, from IBM Lotus Symphony. It was developed over several years by over 50 developers at IBM. But it was done behind the firewall, using proprietary VCS, and when imported into Apache Subversion it lost the entire version history. When it was merged into AOO the work was done over several months and developers in a SVN branch. When the work was done it was them merged into the trunk. How is this work credited by Ohloh, which looks at only our trunk? Yup, it was seen as a single commit by a single developer. But the same kind of merge would, in LibreOffice, be treated as thousands of commits by 50 developers. The fact that LibreOffice lacks partnership with corporations who have significant code to offer is a good thing, but naive log-based reporting misses this fact.

Also, do you have any logo questions? Out of respect for the Lwn.net authors and paying subscribers I feel that we should try discussing the parent article at least a little, rather then rehash the same tired old LibreOffice claims. I realize it burns, burns like acid, to hear someone talking about something other than LibreOffice. But it is best to get used to this.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 17:47 UTC (Sun) by shmget (subscriber, #58347) [Link]

"You seem to take offense that we track website content and design changes in Subversion."
Not at all, I take offense are your attempt to, last year, starting to have these count as 'code', in a desperate attempt to boost your ohloh stats.

" That's useful for our internal tracking, and we use tools like Ohloh to make that tracking easier. "
you did not have them under ohloh for the first year of your existance... and, by your own admission, you forgot to update ohloh 'subscription' after your graduation for many months... so much for 'we use tools like ohloh to make that tracking easier'. When you find yourself in a hole... stop digging.

"But it was done behind the firewall, using proprietary VCS, "

I'm sorry ? are you asking for credit for a failed proprietary fork attempt ?

", and when imported into Apache Subversion it lost the entire version history. "
That was a choice not a fatality.

"With AOO 4.0 we have a huge new feature, the sidebar UI, from IBM Lotus Symphony.[..] When it was merged into AOO the work was done over several months and developers in a SVN branch. When the work was done it was them merged into the trunk. How is this work credited by Ohloh, which looks at only our trunk? "
that account for 105 commits in branches/sidebar, prior to it being merged in trunk.

"It was developed over several years by over 50 developers at IBM."
for a patch that boils down to
1007 files changed, 67109 insertions(+), 18482 deletions(-)

Whoaa 'over' 50 devs over several years... for that ?
No wonder you deliberately squashed the change history...

"Novell had been behaving badly toward OpenOffice.org for many years and waiting for the opportunity."

You may consider that was 'behaving badly' toward Sun, but certainly not toward OpenOffice.org... especially when you try to get credit for a _Proprietary_ fork of the same code base at the same time that happened... the later was not behaving badly toward 'Sun' -- since IBM got their consent for an undisclosed amount of money, thanks to Copyright Assignment -- but certainly behaving badly toward OpenOffice.org the project.

"You are still mixing up commits and committers."

No I'm not. 'still' or otherwise.

"So even with a trivial % of work (which proves my point about the not-existent long tail)"
making a assertion does not prove anything, especially in the face of facts (the so-called non-existent long tail is very very real... you may argue about its value, but not about its existence) in any case ohloh numbers shows trailing 12 months and trailing 30 days numbers... and and LO git repo are there for all to look, nothing done behind firewall and thrown over the fence from time to time...

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 18:29 UTC (Sun) by rcweir (guest, #48888) [Link]

You can't have it both ways. You are arguing that we simultaneously use Ohloh for promotional purposes and try to boost the numbers while also arguing that we neglect the numbers and fail to keep them up to date. You seem to want to argue both sides of this.

But the truth is far less sinister than you imply. The data from Ohloh is worth something, but not much, and we spend some time, but not much, maintaining it. That's our business, not yours. We don't quote the data in press releases. LibreOffice does. I know the limitations and weaknesses of the data and use it accordingly. LibreOffice either doesn't know, or doesn't care.

This is just basic intellectual honesty. If you want to compare things then you need to get them on an equal basis for comparison. I've pointed out the numerous ways in which the processes, history, tooling, etc., of the two projects differ to an extent that the naive comparisons that LibreOffice regularly publishes are deceptive and do not give an accurate view of the activities of either project. Nevertheless, I'm sure it is useful for propagandist purposes and that you, and others will promote it for that purpose. But let's call it for what it is. If you had a product then you'd lead with results.

In-process metrics have their place, of course, but even then you need to be careful. For example LibreOffice is always talking about how many bugs they fixed, but not how many they added. I've looked at many of the bugs and they simply never existed in AOO. Similarly, they quote great performance improvements. But the one I looked at was simply fixing a performance regression they had introduced in a previous version. So 50% performance improvements and bug fixes are promoted, but that is just marketing, not reality. In the end those who pay for the LibreOffice developers will expect results, real results, not just bogus statistics.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 19:09 UTC (Sun) by shmget (subscriber, #58347) [Link]

"You can't have it both ways. You are arguing that we simultaneously use Ohloh for promotional purposes"
No I'm saying you tried to make ohloh number look less bad... bit that you bragged about it... and it is not an 'argument' it is a fact that you decided to add your internal wiki and website to ohloh as part of your project in June 2012.

"while also arguing that we neglect the numbers and fail to keep them up to date. "
Again that is not an 'argument' that is a fact that you yourself expressed earlier.

Beside wanting to mislead people and being bad at it is not incompatible, and clearly your goal was not to 'look good' on ohloh, you know you can't, but to fudge the numbers and make them harder to use... to then try to discredit them. Which is exactly your entire line of argument.

"I know the limitations and weaknesses of the data and use it accordingly"
Of course you know, you are doing your darn best to make them as hard to use as you can.

"LibreOffice either doesn't know, or doesn't care."
Libreoffice does not fudge with Ohloh. We did not add our website or wiki to ohloh 'sources', heck not even our dev-tools repo, contrary to what you did. We do not do 'over-the-fence' development, and yes, we credit the authors of works, regardless of who they are.

"This is just basic intellectual honesty"
you talking of intellectual honesty is like Fox News talking about 'Fair and Balanced'

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 12, 2013 20:00 UTC (Sun) by efraim (guest, #65977) [Link]

You can please keep your politics out of LWN discussions. I try not to bring mine. And as an impartial observer of AOO vs. LO debate your attacks on AOO don't inspire confidence in LO. I would not expect this from a person working for a successful project sure in its future.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 19, 2013 8:20 UTC (Sun) by jospoortvliet (subscriber, #33164) [Link]

Fully agreed. By now I changed my mind about Apache oo vs LO - and not in favor of the latter. Mostly because being unreasonable while arguing makes me doubt the arguments.

Results of the Apache OpenOffice 4.0 Logo Survey

Posted May 19, 2013 17:31 UTC (Sun) by jubal (subscriber, #67202) [Link]

Ah then it's incredibly easy to convince you – it's just enough to start challenging the very point someone wants to convince you about.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds