mmcgrath
[Recent Entries][Archive][Friends][User Info]
Below are the 20 most recent journal entries recorded in the "mmcgrath" journal:[<< Previous 20 entries]
|
01:44 pm
[Link] |
Volunteers, incentives and Fedora updates
So I've been thinking about our whole "updates problem" which to me can be summed up by there's just to many of them and some probably shouldn't have been done (thinking major update changes during a stable release). I myself, am not a very good packager but I am a volunteer packager. No one at Red Hat has ever told me to maintain any packages so I've picked which ones are important to me and the ones I use for $DAYJOB. It's important to establish this relationship when talking about incentives.
When I talk about incentives I'm not talking about payment or cash money of any kind but more what processes are in place to allow me to do the right thing for our users. Obviously all of us have at least some incentive to produce a good operating system. No one is here to make a bad operating system or make Fedora worse. So it's a safe assumption that when Fedora is bad, we all take a little less pride in a release.
So back to updates. When I, as a fairly bad packager, need to do an update, I can do it pretty easily in rawhide. Just build it and it becomes available. At some point in time that build becomes F12 and now my F12 and F11 packages are out of sync. This makes my packages more difficult to maintain as they get out of sync I have to track what patches go where, sometimes even track multiple different releases. From my point of view, I actually have an incentive to keep all of the Fedora releases for one of my given packages at the same revision of the package. This is especially true for a package that may be EOLed and which I have to maintain my own security updates, which is what people pay RHEL to do but no one pays me to do.
Additionally, when I get an update ready I have every incentive to do all releases at the same time. It takes time to make a build, do an update and release it. The current updates system, while designed with more stability in mind with it's karma system, has raised the barriers to do updates. If I build all the releases at the same time and push with bodhi at the same time, from my view it's easier. Particularly when compared to having do and track each one individually. This in itself isn't a problem unless it's combined with the issue above by keeping all of my releases at the same package level which ultimately produces more updates for each release more often.
Also, if I am using F-11 today but I need/want a newer version of some package I'm using. There's little incentive to keep me from doing that. F-11 is stable and what I use at $DAYJOB, the new package fits my use case and as the packager I have access to update it. Then I get my newer package on a stable release and don't have to wait for the next version of Fedora. As an individual packager, I just don't have much of an incentive to do the right thing for our users on my own. If you multiply this by all or some of the other packagers we have and it becomes a problem.
So how to fix it? The board has been discussing this very thing (and will likely punt some things to QA, releng and FESCo for their thoughts). I think there are some technology advances we could make. The updates system does have a lot of the functionality we need but as mentioned above it's not enough. On the technology side, making cvs(or some other scm) less painful to use while keeping different packages at different version levels would be a good start. Also continuing the whole "make BLAH" functionality from our cvs checkouts better has been very welcomed by me. Perhaps I should submit some patches of my own :)
I don't, however, think there's a totally technology based fix for this. We also need some rules in place and people to enforce those rules wrt stable releases. Probably not for every package in the project, maybe just the critical path packages. We may be the only major distribution that does not have an enforced pushing to stable policy in place.
Combined with Jesse's recent rawhide proposal to give the more experimental rawhide a place for any new package while keeping the next version of Fedora more stable, I think we can satisfy most of our use cases and make Fedora more stable. It'll make things more difficult then it is now but we'll have a better OS as a result.
(Note: Made some minor edits)
|
|
10:35 pm
[Link] |
Dear Media
Don't use the term "cloud computing" you don't know what it means. Most technologists can't even agree on the term so it's best to just stay away from it.
|
|
11:30 am
[Link] |
What Is Fedora: Part 2.
More from: https://www.redhat.com/archives/fedora-advisory-board/2009-October/msg00012.html
> (b) Set some broad goals for where we want the Fedora Distirbution to look
> like a few release from now--say when Fedora 15 is released. What should
> those be?
>
1) An easy automated way to provide tests and the answers to those tests.
IE: QA, targeted metric for some given configuration.
This is a "wouldn't it be nice" thing. I have no idea on plans for implementation. But once our automated QA system is setup it would be nice to further broaden it to other hardware platforms and users to do tests, sign off on things. Also provide a way for developers to run tests.
2) More ISVs and start-ups have packages in the distribution.
There has been a stronger push towards this in general and I'd like to see more success in this area if that's even possible. This can be done through better education of upstream providers to use free software licenses. As well as providing one on one help with packaging.
One such metric that was done recently was to try to estimate the number of monitoring solutions installed in the Fedora realm. This can be used to contact other monitoring ISV's (like zenoss) to show the value in providing this software and giving concrete numbers around the possible market for some given software type. More software options in Fedora is a win for everyone.
3) More Architectures. "Does Fedora run on $TOASTER?" Yes.
We're doing so much leading wrt software anyway it seems silly we're not leading this field as well. Secondary architectures is starting to take off. Hopefully by F15 we'll have a rich set of different devices Fedora will run on. We're also starting to put forward additional resources for this type of distribution which should take some load off of the individual teams.
4) A better reputation as a place to bring new ideas to be tested and
presented as well as a better acceptance that failure of a new idea is not
a bad thing.
I just really like the idea of Fedora being a lab for people. If the kernel guys need to check something, come here. If Dell wants to start shipping a new free software bios updater. Start here. Some new startup has a killer app? Start here. We have the expertise for distribution integration, have seen just about every pitfall there is and I'm positive we can handle any pitfalls we haven't seen. This facility in general can help move the industry forward at a much quicker pace then currently.
5) By F15 I'd like to see a killer virtualization management system in
Fedora. What we have now is a lot of disparate tools. All of which are
getting better, none of which are on the level with the likes of vmware.
I'm knee deep in virtualization about every day doing this and that. The stuff Fedora offers in the virtualization world is just not that great right now. Especially if I'm at a company that doesn't care about free software. There's lots of stuff already out there, not only that but Red Hat is working on several of their own solutions at this moment. I'd hope by F15 we will have picked one, properly integrated it in the distribution and added whatever features we need to be a do-it-yourself 'cloud' (eww) distro.
|
11:03 am
[Link] |
What is Fedora: Part 1.
As promised - https://www.redhat.com/archives/fedora-advisory-board/2009-October/msg00012.html
> (a) Define a target audience for the Fedora distribution (or maybe narrow the
> definition to "default spin")--without a clear target audience for our product
> there is a lack of clarity around when we are "done." It also makes it
> difficult to make decisions about release quality and release composition.
>
Experienced users and people that wish to aid in leading our industry
through contribution, experimentation and science. (inventors,
tinkerers, hackers)
I think Fedora should target these users for several reasons.
1. It plays to our current strengths. We are already surrounded by these people. I have never in my life worked with such bright and capable people before. I'm not just talking about the various technologists, but in our ambassadors, design team, docs, etc. They're all good people and many of those teams need more good people.
2. I think these types of users are the most likely to be converted into contributors. A cycle that helps protect Fedora's future by ensuring fresh people are always coming around to take the place of new needs and to replace those that have left.
3. In being first. Fedora regularly overshoots it's means. I don't intend this to be in a bad light. We slipped again with the F12 beta. We slip several releases for whatever reason. At the end of the day though, what we tried to do and what we were able to do in the time allotted, didn't pan out. Experienced users understand this and for features that they are about are more likely to contribute or work with us on specific problems.
4. Aside from Gentoo it doesn't seem anyone is really focusing on this group of people. And even Gentoo seems (IMO) to be doing it in more of a power user way providing a distro for power users, not a place (Fedora Project) and distro.
5. Many of these users also work in enterprise environments where they'll be using CentOS, Scientific Linux or RHEL. It seems natural they'd want to stay in the same family of distributions and I suspect it is where several of our contributors and users come from anyway (that's where I initially came from).
|
|
01:27 pm
[Link] |
DNSSEC
So nb's been working on dnssec for Fedora. Still some work to do but it's mostly working:
$ dig +dnssec +multiline -t ns smolts.org @ns1.fedoraproject.org
smolts.org. 86400 IN NS ns1.fedoraproject.org.
smolts.org. 86400 IN NS ns2.fedoraproject.org.
smolts.org. 86400 IN RRSIG NS 5 2 86400 20091105171909 (
20091006171909 55008 smolts.org.
BXRjrg5TsFpH7wOkCjvZPV9NpBIwjpHoN8VoP57yUV+D
p/jiY3a1N1QgaTui2nBq1gG8zKZGu3X0DMGP9n6L4YIJ
dJ8tem5oLKk8WYiQOyPeJBiC/atLLUhJ/y2N0xkT2lno
cVsdk4gNI1J9D22EcCpOPKFn2ucATRYfhjQp8/8= )
Hopefully we'll have the final 10% done soon.
|
|
04:01 pm
[Link] |
PXEBooting graphic screen with baby
We've all been there. You're in your basement getting ready to pxeboot something, it comes up. Ugh, what a dull text based install. What you really want is a picture. At this same moment, you realize you've been getting way too much sleep. Well my friends, I've discovered the answer. Put a picture of your baby up there every time you pxeboot. So here's what I did. (note, the below assumes you have a pxebootable environment already set up)
1. Find a partner willing to make a baby with you
2. Make baby.
3. Take a picture of the baby. (If you skip steps 1 and 2, I think you can take a picture of someone else's baby. Though this is untested)
4. Save this picture so that it is a 640 pixels wide gif.
5. Convert this picture so that it is only 16 bit.
sudo yum -y install ImageMagick
mogrify --colors 16 baby.gif
6. Convert this gif to a ppm
sudo yum -y install netpbm-progs
giftopnm baby.gif > baby.ppm
7. Convert this ppm to a lss file so syslinux can read it.
sudo yum -y install syslinux
ppmtolss16 < baby.ppm > baby.lss
8. Copy this baby.lss file to the same location as your boot.msg file (usually /tftpboot/)
9. Ensure your boot message is being used and has a "display boot.msg" or whatever your msg file is. Then in that file add the following line:
^Xbaby.lss
Note: that is a control character, not "^" followed by an "X"
Then on your next pxeboot, you should see this baby come up every time. Problem solved!
|
|
11:21 pm
[Link] |
CentOS
So the guys that do CentOS noticed a risk and are working to take care or mitigate it. Suddenly the people that know just enough to be stupid start getting all in a tizzy about it and all the sudden "CentOS is dead"? It's completely irresponsible to be spreading such FUD about a community.
|
|
07:02 pm
[Link] |
Search Domains
So, I was reminded again recently why search domains are a security consideration. I've used search domains in the past as a convenience. You make "fedoraproject.org" your search domain and you can just rsync cvs:: if you want. Nice, convenient.
Well, Someone out there (we think) recently added fedoraproject.org to their search domain. Then they installed func. As a convenience, some apps like func and puppet, try to auto register. Well, we've got this guy regularly registering with our func server. If you don't know what func is, his box is basically asking us what it should do. It's hard to not accept his registration and mess with him. You start thinking about all the fun you could do.... and then not do it. Anyway. It's also why stuff like this:
http://yro.slashdot.org/story/09/07/09/1811249/Comcast-DNS-Redirection-Launched-In-Trial-Markets
Is a pretty big deal and not to be thought of lightly. Stay smart out there, be careful of your search domain.
|
|
01:33 pm
[Link] |
Math Poop
Yep... Math poop.
|
|
10:44 am
[Link] |
Why didn't you vote?
I count more people on Planet Fedora then we had people vote in the elections. We've got over a thousand packagers, but only around 300 voted in FESCo? My question is why? My other question is; How can we get you to vote next time?
|
|
11:25 am
[Link] |
Go. Vote.
It only takes a minute.
https://www.redhat.com/archives/fedora-advisory-board/2009-June/msg00025.html
|
|
04:45 pm
[Link] |
AT&T customer service?
It's rare, and I mean REAL rare for me to write anything good about any company with more then 3 employees :). So brace yourself. Here's the scenario, I've been an AT&T customer since I was in High School. Recently I moved and the cell service where I live is terrible. I was getting about a 50% call rate and a majority call drop rate. It was just not usable. And I tried, oh boy did I try.
So I switch to Tmobile, got a G1 and things were good. Well today in the mail I got final bill from AT&T with some fairly triple digit cancellation fees. So put my guard up, prepare for a battle, and call AT&T. I was on the phone with someone for maybe 45 seconds, most of which was spent trying to find my account since it had been cancelled.
She then proceeded to volunteer her name - First and last. I've had people refuse to even give me their last name when I ask for it, she volunteered the whole thing. Didn't even ask.
She then put me on hold. About 3 minutes later she got back on the line and said "I can see all the notes from some of your previous calls on this issue and was able to cancel the account without any cancellation fees." She then went on to apologize they didn't have service in the area and remarked about how long I'd been a customer. I said thanks, she said thanks, and that was the end of that transaction.
Completely not what I was expecting, considerably shorter time then I expected it to take and even now that I'm no longer a customer, I'm completely satisfied with AT&T. If you know anyone that works for AT&T, pass this along.
|
|
12:29 pm
[Link] |
Memcached Plug
I recently started using memcached and I have to say I'm quite impressed with it. We'd used it sort of on a trial basis in Fedora Infrastructure for some time (Thanks Ricky). But after playing with it a bit with python, and after moving all of the wiki to using it, I'm even more impressed then ever.
Why?
Because memcached doesn't do much, it's very simple. But what it does do, it does very, very well. After just 15 minutes trying to figure out how it worked and I could predict how it would behave, what it would do if I did X, Y or Z. This made it easy to code for. But even with as easy as implementation was, I've been extremely impressed with how it performs. Completely predictable. I have yet to be able to make it fallover. It's the sign of a well written, well designed application.
Memcached is "missing" things. Security immediately comes to mind. Redundancy as well. But neither of these things are memcached's job. Your app and deployment has to be memcached aware. You need firewalls, you need your application to update the cache, deal with failure scenarios, etc. Memcache won't solve all your problems, after all, it only does one thing; caching of data, but holy crap does it do a good job of it.
http://www.danga.com/memcached/
|
|
04:42 pm
[Link] |
Anyone video conferencing?
My folks want to try to do video conferencing... Anyone else doing this? What hardware are you using? Anyone have any non-computer suggestions here? Set-top box type thing?
|
|
04:42 pm
[Link] |
Community
I live and breathe Fedora and as a result have gotten to know it's community quite well. I get this community and I generally think it gets me. It wasn't until I saw another, thriving community that it made me think about things.
What is this other "thriving community?". It's Valve's users. It's not a huge secret that when I'm taking a break from breaking things in Fedora, I sit down and bust some heads in TF2. I wish it were free software, I do. I'd sleep better at night. But at the same time, Team Fortress 2 is probably the most fun there is to be had online right now. What does this have to do with community? Well, Valve (makers of tf2) has actually listened to it's user base and repeatedly released updates to it's games. The recent being the sniper/spy update. If you follow the game at all then:
http://tf2.com/
Will make a lot of sense to you and the recent "meet the spy" video will be very amusing to you:
http://www.teamfortress.com/spyupdate/day05_english.htm
Even though they don't give the code away they are more open then most companies and they actually do listen to the people that use their stuff and if their audience is me, then they sure seem to know what I like to play because every game they release is just amazing.
So valve, my hats off to you. Keep the entertainment coming!
|
11:16 am
[Link] |
Kindly FAS reminder
If you've not logged in and verified your Fedora Account System - https://admin.fedoraproject.org/accounts/ stuff is correct, please do so. If you've changed jobs, email addresses or phone numbers, make sure they are up to date.
I'd also *strongly* suggest you use your personal information for the email address, postal address and phone number because it's a way for us to verify who you are after you've left that place. If you work for someone and suddenly have to leave and your account expires, you might be out of luck and have to create a new one.
And yes, our privacy policy is legit, we won't be giving your telephone number and stuff to others, we might call it though if someone shows up on IRC claiming to be you!
|
|
12:40 pm
[Link] |
Fun
1 Sledgehammer. $20
1 crowbar. $15
1 sawzall. $17.50 (rental)
1 unwanted hot tub. $no idea
1 awesome mothers day of destruction? Priceless
|
|
09:05 am
[Link] |
Where are we?
Why is Fedora so full of fail with stuff like this?
http://www.computerworld.com.au/article/300094/pictures_highlights_from_2009_google_summer_code
* The Linux Foundation - Check
* Firefox - Check
* OpenSuSE - Check
* Drupal - Check
* KDE - Check
* Debian - Come on check
* ASF - Check
* Gimp - Check
* Git - Check
* Gnome - Check
* Joomla - Check
* Python - Check
* PHP - Check
* Ruby - Check
See Fedora on that list anywhere? I know we're doing SoC projects. What exactly are we the leader of?
|
|
12:28 pm
[Link] |
Oh comcast
While transferring my service today:
Daisy: Have you considered the Triple Play? The Comcast Triple Play gives you Internet, phone and cable together in one package for one low price each month for an entire year/ 6 months when you sign up for all three. You'll get Comcast Digital Cable with Channel 1 On Demand, Comcast High-Speed Internet that's way faster than DSL, and Comcast Digital Voice with unlimited local and long-distance calling. Would you like to sign up for this?
Michael_: naw, we don't need phone at the new address just cable and internet.
Daisy: I appreciate the hesitation, but this is exactly what you need since you are moving. With just one bill and one relationship to manage, the Comcast Triple Play delivers simplicity. There's no more cost-effective way to tap into Comcast's full range of services. This would be a perfect time to add this service at a discounted rate to your account. Would you like me to do that?
Michael_: I mean we literally don't need a phone in the house, haven't used land line phones in over 3 years.
Daisy: are you sure?
Michael_: no means no.
|
|
07:44 pm
[Link] |
The problem of leadership
So this is a good example of the image problem I think Fedora (the leader) needs to overcome:
http://lwn.net/Articles/325761/
This person 'is pleased with the arrangement'
He understands "his users" benefit from it
Yet words it as if we have to trick our "unwitting users" to alpha test his software.
So why on earth did this guy feel the need to put Fedora in anything but a favourable light? We're happy, our users are happy, he's happy, his users are happy. It's ok to follow the leader, really.
|
[<< Previous 20 entries] |
![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
mmcgrath's journal