Slashdot Mirror

IIRC .deb doesn't have some of the nice error checking that rpm has, like signing. Also compare files on a system to the originals in an .deb to check for tainted files (though while looking I see that debsums(1) might do the trick.

Other problems is the PITA it is to do an non-interactive install of debs. One of the biggest bitches I hear about debian is that when doing an initial install, and you've gotten to the part where packages are installing it goes something like this:

*install*
*install*
*install*
*ask question via debconf*
*install*
*install*
*ask another question*
*install*
*install*
*install*
*ask same question again*
*install*
*install*
etc...

Also .deb's inadequate logging is mentioned, as well as keeping the install messages somewhere, or keeping previous versions of a package (what happens when you find out that libfoo is completely b0rked in the latest version, and have to run around searching for a mirror site that hasn't been updated yet. This generally only happens in unstable, but it's still a PITA :)

There were a couple of other features that .deb didn't have, but I don't recall them right now.

Some references and info is here though that's a lot more pro-deb than discussing this exact issue, but there's good info there.

Oh, and before you start flaming, I'm a long time debian user :)

Here's a slightly different thread from DebianPlanet, but apparently this issue has had quite a few ramifications.

But this isn't the reason why I use Debian. That reason is the deb format. It's quite simply far more powerful and more consistent.

Here's a discussion of the issues by a Debian package maintainer

But deb format maintenance requires a certain package developer/maintainer culture that might tend to clash with the requirements of the kernel/base developer/maintainers, which are rather different from those of general package maintainers. And Debian-based systems can already deal with rpms, no problem, using alien. So using rpm for standardized base is a no-brainer.

The whole notion of mebibytes is an ugly, illconsidered, and overly specific hack designed to fix a real, but by no means debilitating, problem. I made a similar post, anonymously, to debianplanet this morning, where I first saw this subject discussed, and only just now got around to checking out slashdot. I should warn you up front the the following is fairly opinionated rant, and probably represents a rather unpopular opinion to boot. You have been warned. :-)

The thing that really annoys me about the whole Megabyte/Mebibyte thing is that the entire standard nomenclature is an ill-considered, quick, dirty, and above all ugly hack addressing an admittedly legitimate problem (Mega meaning 1^06 or 2^20).

Their hack addresses only powers of 10 and powers of 2, which are a subset of a larger question: nomenclature for abitrary (integer) bases. Worse, it mixes the two together in a misguided effort to get one base's representation to approximate the others, despite the fact that the two bases are in fact quite different!

Why is this so stupid? Well, aside from the internal lack of logic (and, I have to say, profound lack of elegance), let's suppose, for example, that in ten years we begin finding more widespread use of balanced trinary systems , or some other hitherto unforseen base. Where's our nomencalture now? Of completely no use, and requiring us to invent a new wheel, yet again.

A far more reasonable approach would have been a subscript denoting the base, with the default being base ten if no subscript is present (i.e. defaulting to standard metric nomeclature). E.g. M(sub)2Byte would be 2^6 Bytes while M(sub)10Byte = MByte = 10^6 bytes. A M(sub)3trit would be 3^6 trits, and so on.

One will immediately notice that what we consider a (base-2) Megabyte is not 2^6, but rather 2^20 Bytes, or 64 vs 1048576 bytes. Well, they want us to learn a different nomenclature anyway, so why not at least make it logical. If Mega always means to the power of six, regardless of base, then we have a rational basis for our nomenclature. Yes, it would take some getting used to, but I would argue it would be far less painful getting used to something this logical than to adopt the use of "mebibyte" in our daily language. YMMV of course.

This ugly hybridization of base-10 nomenclature with base two numerology they are intending to replace (admittedly equally bad) common usage with is both illogical and unnecessarilly specific to one problem set. If we're going to be making up new (and apparently stupid) terms like mebibyte, then lets at least define mebi to represent a power of 20, or better yet 21 as it would then follow exa by an additional power of three, as every other prefix above kilo (and below milli) does. Or better yet, pick a name that doesn't start with the already (overused) 'M'.

Does anyone else see the advantage of this? We have just extended our available nonemclature for all measures, in any base, in a rational, extensible, and fairly scalable approach. Yes, to our base-10 minds we may feel uncomfortable with the small size a Megabyte really is, but I would submit that that is no greater a psychological barrior to overcome than the use of really stupid, childish, and annoying terms such as "mebi," and a heck of a lot more rational to boot.

Of course, this idea came from one person, spontaneously, on a Sunday morning, who (at the time) hadn't even had his coffee yet. Give a self-appointed committee time enough to dumb it down and who knows what hideous form it would then take...

Regardless, since you're obviously not interested in giving accurate statements about the current state of the Debian projects, its software, or its developers and users, why don't I throw out a URL for those that may be interested. You can simply over look this:

• www.debian.org
• www.debianplanet.org

1) I had woody on all my servers since febrary, maximum uptime, not problems with the packages, etc..

2) There are Woody ISO's
Go see http://www.debianplanet.org/debianplanet/article.p hp?sid=207

He's been quite active lately, applying to be a Debian developer and all... he must be planning something big, I don't remember him being this "visible" for years.

Nice job, Eugenia. That was word for word from an article on this mornings Debian Planet
We don't usually see such blatent plagurism, but hey. it works ;)

Nice job, Eugenia. That was word for word from an article on this mornings Debian Planet
We don't usually see such blatent plagurism, but hey. it works ;)

My question is, how is apt so much better than, say, Red Carpet on a RedHat box? It works out dependency issues, just like apt does.

For technical differences, see this article. One of the biggest ones is that debs have a higher degree of dependency granularity than rpms. As well as Depends: and Conflicts: they have Recommends: and Suggests:.

Debian also has a carefully thought out packaging policy. And in Debian, everything is a true Debian package. There is no contrib. So a bug against the package is a bug in the system. This mindset makes a big difference to the quality of the distribution.

You might also be interested in this discussion on this theme.

i completely agree with you on your view of potato. some packages, like the entire gnome distro, are outdated. you could change you sources, and pull down gnome from woody or sid. not that i would recommend that, of course. undoubtedly, they would require some new version of glibc, etc. i run sid on my daily workstation (not my server or firewall) and love it. it's not as broken as one might think. there are days when some packages won't update due to a dependency not being uploaded yet, but it always seems to sort itself out in a day or two. debianplanet does a good job of letting people know about the packages that are really broken (X was the other day). i switched to sid from redhat, and there's no turning back for me.

On a related topic, there's a pretty good discussion of the relative merits of .deb and .rpm (the file formats, not the programs) here. Seems to come down mostly in favor of .deb, which isn't surprising - who would you expect to read Debian Planet?

From a debian planet article:

The third in as many prominent efforts into releasing ReiserFS Debian Install Disks has been made by Zoltan Kraus. Though Zoltan has improved on his predecessor, John H. Robinson, IV's previous work substantially. Most notably Zoltan's disk are using 2.2.19 and ReiserFS 3.5.32 accompanied by the latest reiserfsprogs. His disks are at http://debianboot.digitaltux.com/.

They are also mirrored at http://www.markybob.com/zoltan/2.2disks In an email from him, he claims that "They have been tested and work perfectly on both desktops and laptops." Other patches that have also been included in the disk set are: Raid Patch 0.9, Raid1 ReadBalance, Hedrick's IDE Backport, Solar Designer's Secure Linux Patch, among other miscellaneous performance related patches.

In what could be a coup for Zoltan, he may become one of the first people to release a 2.4.x Debian install disk for public consumption as well as a set of XFS Debian Install disks which will be keenly sought after by many cutting edge Debian users. "I plan to make a set of XFS disks when they bump up their CVS to kernel 2.4.3 and a set of ReiserFS 2.4.3 disks when I have time to debug it." Seeing that 2.4.3 is out now, be on the lookup in the next few weeks.

Maybe he was waiting for XFS 1.0 - should be real soon now at any rate.

DebianPlanet

DebianPlanet

DebianPlanet

What's the difference between this and Debian Planet. Seen it in the topic on OPN. I figure this must be the eh? Offical help site? Ideas?
--