First off, sorry to see you go Alon and I appreciate the work you did for Gentoo.

I empathize with your views that you expressed in your mail. These are difficult problems and it is not a quick change to fix them and you (nor I) may not agree with the solution. I think Gentoo has many goals; however they are poorly expressed and because of that there is no way to essentially ‘meet’ them.

The problem we previously had was that ‘everyone thought they had a say’ and that appeared to not work. The council is attempting to modify that view somewhat and it is a slow process. You can’t just elect some dictator guy to make arbitrary choices and set arbitrary goals and metrics. You need a critical mass of folks who are willing to dedicate time to meet those goals. You must also be willing to accept losses of developers who disagree with this new vision.

I know many developers who left because they thought change was impossible. I used to think that way. One of the things that getting a real job has taught me is that changes like these are not impossible, they are just not easy. You need folks dedicated to actually solving them and not just talking about it. People to set goals and actually sell them to the community as worthwhile and convince others to work towards them.

Maybe we used to have that calibur of person; maybe we never did; maybe we never will. But you are correct in one thing. Someone has to try. If everyone fears making the necessary changes than Gentoo will stagnate and probably die.

I actually hate REST because all of its supporters claim it is ‘Teh awesome!’ but so very few can tell me why in any sort of convincing manner. ‘It doesn’t have all the overhead of SOAP!’, ‘It uses HTTP!’…these are not convincing arguments. Anyway I figured I’d give it a try.

gStats now has no real webUI. Most of the data retrieval functionality is written (aka HTTP GET) using old data in the cache so you can do queries like /api/cp/sys-apps/portage and it will return some XML that describes sys-apps/portage’s installation history. Also available is /api/arch/x86 which will return some XML describing x86’s installation info.

I am working on writing POST methods; although I noticed I have a slight issue. Users have hosts with 100’s of packages installed. REST is kind of limiting in this regard; right now I store arches and cps (category/package strings) and via the rest interface you can POST new arch data (not really allowed at this point as it would overwrite old data in my reckoning) and you can POST new cp data (being worked on). But the cp POST model right now kind of takes a cp as a key; maybe I should drop that requirement.

Currently:

POST /api/cp/sys-apps/newpackage
some XML describing a single cp named sys-apps/newpackage

New Idea:

POST /api/cp
some XML describing multiple CPs


The tree is too big (said while watching emerge –sync run). Copying all this data down over wireless sucks and perhaps I will devote a tiny slice of time to making a subtree (of packages I actually need) and not sync the other 12000 packages….

Pkgcore-0.4.7 is in gentoo-x86, put there by me because ferringb’s normal commit monkies were not available and lo’ I still have gentoo-x86 access!

I switched from fluxbox to ratpoison. I barely use a mouse now! (ok probably a lie, I am using one right now). However I think it is a plus overall; Yet Another Set of Keybindings is annoying though

There is little information as to exactly why (certainly there have been past allegation of improper behavior by these developers and repeated complaints to council@gentoo.org).

I have made a formal request to the council for more information regarding this matter. I think we should strive for transparency in matters such as these. I applaud the decision to act in this case although I hope the council will provide more information about this decision soon.

I really neglect this blog

That being said; work has been super busy lately. I hope next quarter I have more time to embark on fun life things. I just got back from a trip to Las Vegas and it was an enjoyable weekend, albeit one with a lot of walking. During the trip i read the book verison of Jumper. Lets skip the part where the film was horrible and get to the part where the kid steals millions of dollars from a bank, doesn’t get caught, and then moves on to spend it in what I can only describe as a ‘fairly reckless mannor.’

My point in bringing this up is not to rob a bank (unless you enjoy jailtime) but that it is…tempting a bit to live beyond ones means. My trip to vegas was a fun time and after tallying up the costs…I wondered how much I could realistically spend doing crazy things like ‘going to new york for a weekend and seeing some shows’ or other things like that. I would have never thought of doing something like that a year ago.

Obviously spending tens of thousands of dollars on silly trips every couple of months is no way to save money and in the long run it is prudent to have a budget and a plan for saving cash. Right now I am doing ok there but I think as I consider things in the future (3-5 years) I wonder at what I can do to get to that sweet spot faster. It would be great to not have to stop working (which would be boring I think) but perhaps to not have a ‘real job’. This thought process of course, being triggered by a co-worker who is taking a year off (yay stock options) to do random crap. Being that I’m single, no car payment, no pets…etc… now would be the time to do something crazy.

And so I join the millions of Americans who are trying to think up ways to make money

I decided to get there. The gentoo stats code that was previously written is fine but I thought would not suit my needs.

My goals were:
* Finish in under 14 days.
* Non-complicated
* Easy
* Fast enough

I came up with gstats.

gstats is:

* a multi-threaded python webserver (from Paste).
* that takes in POST requests of ASCII data (Colubrid and Routes)
* Has a data injection thread that injects uploaded data into the in-memory cache and then serializes the cache to disk.
* Has some views to view the data in the in-memory cache (Mako Templates)
* uses the new dev-cpp/gflags

Currently gstats only supports ARCH and CPV data (so who is using what packages on what arch).
The POST format is ASCII and is still being iterated on. I plan on running this in a public ‘alpha’ in a bit.
The Mako templates need work.

There is no SQL involved. The disk format is:

uploaded/ - contains files uploaded by clients.
procesesd/ - contains files we have added to the cache.
archived/ - contains files that are old.

The queries are not complex enough to require anything else. The in-memory cache probably has some nasty memory requirements but is probably OK for now.

If you are a dev you can git clone git.gentoo.org:/var/gitroot/gstats.git

Otherwise you are SOL until I push the code to anongit.gentoo.org, which will happen when I get time.

I was going to have a bunch more content; but it will be on www.gentoo.org as soon as I can get some feedback on certain items. This was going to be a part of that news item but it was deemed too editorial (drawing conclusions about time constrains and motivations) and so I took it out and moved it to my blog where it is appropriate to make editorial comments.

I wish Daniel the best and good luck in his efforts. He is a great guy and I’m disappointed that he seems not to want to consider helping Gentoo as a typical trustee or council member; both teams on which he would most likely be most welcome. He raises valid points about the current state of Gentoo and there are threads on both gentoo-nfp and gentoo-project on how to solve them. For the avid users out there; I’d like to remind you that the distribution cannot move forward without you and the time you volunteer. If you wish to help out there is the completely out of date staffing needs page.

Here is a hint: Not everyone gets along all the time. Shit happens, people say mean things, grow a pair (balls for men, I guess breasts for women, who knows*).
That being said, we tend to have to work with people whom we don’t always hold in the highest esteem. Dealing with your fellow man (or woman) is an important life skill
and I find it difficult to believe that we have a large faction that just lacks the ability to do so.

Realize then, that some people are just pricks who want to tick you off and are well versed in how to do so.
Realize then, that some people are just hungry for power and are well versed in how to gain and keep it, and also well versed in how to use it against you.
Realize then, that some poeple are quiet fellows, more concerned about their packages within Gentoo than Gentoo itself. Realize that you should not alienate these people, as they give Gentoo the girth it needs, the wideness of packages, to make it useful to so many.
Realize then, that some people are just trying to help; they could care less about the thing you are arguing endless about, they just want the arguing to end.

Realize then, that you need you deal with all of these people in order for Gentoo to continue. Trying to get them fired or trying to make them quit does not serve Gentoo. Gentoo needs its people, skilled people who know what they are doing, have experience in the area, and have served us well in their area in the past.

I don’t get why everyone’s personal differences preclude them from being civil, from working together, from acting like mature persons. It’s always ‘the other guy did it’, ‘it’s not my fault’…etc…etc.. Seriously man, chill.

* (Breasts must not be hairy)

Not that they should, because the code fucking sucks. Now that is coming from me (and it’s only meant to be half mean). It’s hard to criticize code that works, it runs much of gentoo (certainly portage is a monstrosity that many of you actually use successfully on a daily basis). I have crazy high software standards. That being said, I also tend not to write a ton of software, most of it small and single service. Writing good complex software is a difficult process. (more accurately, it’s time consuming, and I have a very small attention window)

That being said, a bundle of cgi scripts in python that barely work isn’t exactly a shining example of open source or software in general. But it ran packages.gentoo.org for years and we all (myself included) appreciated its inherit usefulness. That being said, as I stare at the code in cvs and wonder if I should spend some time on it..my mind keeps latching on to a rewrite. I really hate patching code that doesn’t work to be code that barely works. There is an inherent thing within me that just screams ‘rewrite this’. But then that has problems too.

You see rewrites have this common thing called feature creep. You can rewrite p.g.o, make it use html templates perhaps instead of print statements Make it use 2 or 3 scripts instead of 18. Move the db/sql stuff to an abstraction layer (I normally don’t like abstraction layers, but this app is pretty simple.) But then I get to thinking…Hell we can use pylons or django! We can store more data! We can use a python -> XSLT bridge to generate output that is similar to the rest of www.g.o! We can do all kinds of things. But we probably shouldn’t.

Most people just want the service back up and they will take a few monkey patched python scripts to do it.

The application then stores the diff.

Gentoo Users and Developers can critique the diff, offering suggestions on how to improve the code as well as advice
in areas were the code is crap.

This may or may not be useful. Certainly non-developers may be miffed by the fact that many developers won’t ever read their
comments or may never incorporate suggestions into the tree.

However I think it would be an interesting application to write as well as a way for learning developers to get feedback from other
experienced ebuild developers and users.