Hi,
Didi and I met up yesterday afternoon and had a chat about creating a web interface to the centos repositories. Something along the lines of repoview, but a thicker, richer experience. So what features would everyone like to see from something like this ? We have a few ideas already, but it would be nice to get feedback on what people expect ( so we know how far off the mark we were :) )
We hope to resurrect the 'pandora' project as a side effect from this.
- KB
On 06/22/2010 02:20 PM, Karanbir Singh wrote:
Hi,
Didi and I met up yesterday afternoon and had a chat about creating a web interface to the centos repositories. Something along the lines of repoview, but a thicker, richer experience. So what features would everyone like to see from something like this ? We have a few ideas already, but it would be nice to get feedback on what people expect ( so we know how far off the mark we were :) )
We hope to resurrect the 'pandora' project as a side effect from this.
IMHO, the most important thing is 'search' something like Rpmfind.net
I used to use google with some tricks like "site:vault.centos.org Index of $PACKAGE_NAME"
Thanks for advance.
On 22/06/2010 14:24, Athmane Madjoudj wrote:
IMHO, the most important thing is 'search' something like Rpmfind.net
Yes, good point. I've been looking at a few different search engines that we might be able to use here. The one thing that we need to do some level of work on is : weights. How do different parts of the metadata get weights to make them more or less significant.
- KB
Yes, good point. I've been looking at a few different search engines that we might be able to use here.
Google ?!
The one thing that we need to do some level of work on is : weights. How do different parts of the metadata get weights to make them more or less significant.
Search priorities (from high to low):
Package name > Provides > Summary (in it too) > Files
I don't thinks that changelog is useful for end-users
On 22/06/2010 14:47, Athmane Madjoudj wrote:
Yes, good point. I've been looking at a few different search engines that we might be able to use here.
Google ?!
We could try asking politely, but I dont think they will give us their indexer / engine to use :)
Search priorities (from high to low): Package name > Provides > Summary (in it too) > Files I don't thinks that changelog is useful for end-users
How about payload contents ?
Also, perhaps not as a search result, but I think changelog is a good thing to have. eg. being able to do : tell me the changes for whatever package provides /etc/blahblah from what I have now, to whatever is the latest in remote $repo; is worth having. So this could be expanded into something like : show me the overall impact that doing a yum update is going to have on /var/lib/mysuperlibfoo/ ( ok, so I admit its a corner case, but still )
- KB
We could try asking politely, but I dont think they will give us their indexer / engine to use :)
I meant Google search API or Google custom search
How about payload contents ?
Also, perhaps not as a search result, but I think changelog is a good thing to have. eg. being able to do : tell me the changes for whatever package provides /etc/blahblah from what I have now, to whatever is the latest in remote $repo; is worth having. So this could be expanded into something like : show me the overall impact that doing a yum update is going to have on /var/lib/mysuperlibfoo/ ( ok, so I admit its a corner case, but still )
I'm still wondering about the tool because:
1. repoview [1] is too basic 2. rpm2html [2] is deprecated :-( 3. rpm.pbone.net tool is not available for download.
...
[1] https://fedorahosted.org/repoview/ [2] http://www.rpmfind.net/linux/rpm2html/ [3] http://rpm.pbone.net/index.php3/stat/2/simple/2
On Tue, Jun 22, 2010 at 9:47 AM, Athmane Madjoudj athmanem@gmail.com wrote:
I don't thinks that changelog is useful for end-users
This depends on your target user audience. For the average home user or basic office set-up you're probably right. I know for the folks doing PCI compliance the changelog is rather helpful when CVE closures are listed. That could be done with a completely separate tool though, and rh already has a decent model/tool for searching CVE listings.
On 6/23/2010 12:46 PM, Jim Perrin wrote:
On Tue, Jun 22, 2010 at 9:47 AM, Athmane Madjoudjathmanem@gmail.com wrote:
I don't thinks that changelog is useful for end-users
This depends on your target user audience. For the average home user or basic office set-up you're probably right. I know for the folks doing PCI compliance the changelog is rather helpful when CVE closures are listed. That could be done with a completely separate tool though, and rh already has a decent model/tool for searching CVE listings.
Is there any chance of making it able to detect and flag conflicts among any/all participating repositories in terms of duplicate but out-of-sync package versions and conflicting file content? This information could be used both by the package builders to clean up the problem where it was accidental and by end users to restrict where their updates are pulled.
On Wed, Jun 23, 2010 at 1:57 PM, Les Mikesell lesmikesell@gmail.com wrote:
Is there any chance of making it able to detect and flag conflicts among any/all participating repositories in terms of duplicate but out-of-sync package versions and conflicting file content? This information could be used both by the package builders to clean up the problem where it was accidental and by end users to restrict where their updates are pulled.
I don't believe so. It's mostly just an xml file output. No idea how the back end is generated really. You can check out the site here ->
http://www.redhat.com/security/data/metrics/
Karanbir Singh wrote:
On 22/06/2010 14:24, Athmane Madjoudj wrote:
IMHO, the most important thing is 'search' something like Rpmfind.net
Yes, good point. I've been looking at a few different search engines that we might be able to use here. The one thing that we need to do some level of work on is : weights. How do different parts of the metadata get weights to make them more or less significant.
Have a look at the work Nikolay has done here:
http://pkg.org.ua/ http://www.lystor.org.ua/projects/personal/pkg-org-ua
Makes it very easy to find packages across the whole EL community. Lots of other nice features too.
On 22/06/2010 15:07, Ned Slider wrote:
Have a look at the work Nikolay has done here:
http://pkg.org.ua/ http://www.lystor.org.ua/projects/personal/pkg-org-ua
Makes it very easy to find packages across the whole EL community. Lots of other nice features too.
that just looks like a skinned repoview. What am I missing ?
- KB
Hi Karan,
I have used 'packages.debian.org' a lot and find it quite helpful. Implementing something similar to this in CentOS would be nice.
It lets me choose which 'distro' I want to search by name. After I find the package, it shows the full depends, changelog, etc.
For example, for php5 in 'stable' I get a list of packages, and if I choose 'php5' I get the following page: http://packages.debian.org/lenny/php5
This shows a ton of information about this, including bug reports against this package on the right hand side among other things. It also lets me download the package straight from here, without having to go searching through a repo. Similar packages, maintainers, external resources are also listed here.
In addition, the page is available in multiple languages.
Quite often someone will ask me whether a certain package is available in Debian stable or not and if it's not I can easily search 'testing' or 'unstable' here and I quickly know that the package is available, but has not made it to testing/stable yet for whatever reason. Then a decision can be made as to whether we should try and get it from testing or just compile it ourselves, or just wait for it to appear in stable, etc.
Just some of my thoughts on the matter. Thanks for implementing this.
Khusro
On 22 Jun 2010, at 14:20, Karanbir Singh wrote:
Hi,
Didi and I met up yesterday afternoon and had a chat about creating a web interface to the centos repositories. Something along the lines of repoview, but a thicker, richer experience. So what features would everyone like to see from something like this ? We have a few ideas already, but it would be nice to get feedback on what people expect ( so we know how far off the mark we were :) )
We hope to resurrect the 'pandora' project as a side effect from this.
- KB
CentOS-devel mailing list CentOS-devel@centos.org http://lists.centos.org/mailman/listinfo/centos-devel
On 22 Jun 2010, at 16:17, Khusro Jaleel wrote:
Hi Karan,
I have used 'packages.debian.org' a lot and find it quite helpful. Implementing something similar to this in CentOS would be nice.
It lets me choose which 'distro' I want to search by name. After I find the package, it shows the full depends, changelog, etc.
For example, for php5 in 'stable' I get a list of packages, and if I choose 'php5' I get the following page: http://packages.debian.org/lenny/php5
This shows a ton of information about this, including bug reports against this package on the right hand side among other things. It also lets me download the package straight from here, without having to go searching through a repo. Similar packages, maintainers, external resources are also listed here.
In addition, the page is available in multiple languages.
Quite often someone will ask me whether a certain package is available in Debian stable or not and if it's not I can easily search 'testing' or 'unstable' here and I quickly know that the package is available, but has not made it to testing/stable yet for whatever reason. Then a decision can be made as to whether we should try and get it from testing or just compile it ourselves, or just wait for it to appear in stable, etc.
Just some of my thoughts on the matter. Thanks for implementing this.
Khusro
Sorry about the top post, I sent that in a hurry...
On Tue, 22 Jun 2010, Karanbir Singh wrote:
Didi and I met up yesterday afternoon and had a chat about creating a web interface to the centos repositories. Something along the lines of repoview, but a thicker, richer experience. So what features would everyone like to see from something like this ? We have a few ideas already, but it would be nice to get feedback on what people expect ( so we know how far off the mark we were :) )
I do not know that it is going to mees the complaints seen in the IRC channel, but a simple 'reverse criss-cross' on exploded package manifests can begin to populate a query database such as at: http://www.debian.org/distrib/packages
Performance may matter, so gatewaying into a rpmlib quiery against a rpmdb [yum can do a variant of this with detail from a createrepo as well] may or may not work
The 'current' 'base and updates' archives are the obvious first candidates to test a design against; no doubt some one somewhere will want all of 'vault' as well, and I would not be inclined to do that
The top page at http://www.owlriver.com/projects/ORC/ is refreshed one a day in a scripted generation process using selected --queryformat arguments against a simple tree of rpm packages; I auto-build that cache as one off lookups are too expensive
-- Russ herrold
Check out freshports.org, they've got a really cool setup for searching the FreeBSD ports tree. I'd look for something with similar functionality.