David Kanter (dkanter@realworldtech.com) on 2/1/09 wrote:
---------------------------
>anon (anon@anon.com) on 2/1/09 wrote:
>---------------------------

>>Huh, funny everything suddenly goes quiet. You'd think >someone would have to be
>>pretty sure of themselves before challenging someone like >Linus (or any other ~20
>>year chief architect of a successful OS) on a position he >clearly feels quite strongly about.
>
>I'd point out that Bill Shockley felt strongly that whites were superior to non-whites.
>That doesn't mean anyone should care about whatever crazy views he had.

OK, but that isn't really an equivalent situation. One could disagree with this character with good reasons to back up that position.

I wasn't saying listen to Linus because of who he is, actually that doesn't add too much to his position. I just find it surprising that people would seem to get so fired up as to jump in with insults while it seems like they actually can't answer any of these basic questions or explain why xattrs are better. OK, I guess the same applies when arguing against Linus or "nobody", so maybe people are just trolling and arguing *because* of who he is.


>Just having a strong opinion doesn't make much difference. The question is one
>of relevance and the extent to which the expertise is one field applies.
>
>I know a lot about CPUs, and I have an opinion that from a competitive position,
>AMD is in trouble. That's a good reason to think AMD is in trouble.
>
>OTOH, if I tell you that based on my expertise in CPUs that I think GM is in deep
>shit, that doesn't really add much to the picture.
>
>Obviously Linus knows a bit about operating systems and processors, but I'm not
>sure how much of that applies to file systems. Perhaps someone could comment on the connectedness between the two...

I don't think state of the art filesystems is Linus's strong point. But the issue isn't really the filesystem. On a modern filesystem with a good xattr implementation, I think they implement them similarly to files, eg. with btrees for data and possibly even with hashes to enhance lookup. But that part of it -- implementation -- is uninteresting. It is not going to be particularly faster than regular directories/files, but probably just because there has been no need for it yet. I don't know of any fundamental reasons why either should be notably superior at the filesystem level.

No, the real issue is the namespace and API level. And from what I have seen, Linus I think does have a lot of experience in that area. But yes either way, it doesn't matter that it is Linus, just the technical points of the argument. And those arguing for xattrs don't seem to be doing very well there.


>That being said, I thought Linus' point about xattrs having to really benefit everyone
>on average seemed to be on the mark.
>
>Simply having an example of where something is helpful isn't useful.

Exactly one of the points I was making as well.