Quantcast

GoboLinux – A Breath of Fresh Air?

Get the WebProNews Newsletter:
[ Business]

My very first reaction to GoboLinux was negative. The underlying idea of taming the Unix/Linux file system hierarchy with symbolic links isn’t new: heck, SCO did that way back with their 3.2v4.0 release, and for exactly the same reasons (see the Software Storage Objects- SSO section at this link).

Most of us in the field agreed that the concept was laudable in theory but failed in practice: too many places where it broke down. See Speaking of symbolic links… for some taste of that. The basic flaw of this sort of scheme is that symlinks are easily broken; there’s no warning that removing /Foo is going to remove access to the entire hierarchy that appeared under it. Oh, of course the things /Foo pointed to aren’t gone, but that doesn’t make things less broken when it happens.

But.. you know, the more I read about GoboLinux (especially the “I am not clueless” article), the less certain I became that this is a foolish Don Quixote effort. Yes, yes: “Those who do not understand Unix are doomed to reinvent it, poorly.” – Henry Spencer, 1987. Yes, yes, it’s not hard to find examples of GoboLinux having broken link problems. Nor is it hard to find people vehement in their opposition. But the split between those who like it and those that do not might be pretty even: this poll at the Ubuntu Forums showed 42.59% liking the concept and only 35.8% opposed (21.6% were neutral). That was only 162 people, but it shows that this isn’t “obviously” bad.

The file system links aren’t the only oddities: neither BSD or SysV style rc scripts are used. Instead /etc/inittab has “/System/Links/Executables/BootDriver” – a shell script that replaces all that. There’s also no “root” name in /etc/passwd: the id 0 user is “gobo”. I’m still not clear on why that was done, but of course it’s easy enough to change (and the docs even speak of perhaps allowing the installer to choose id 0′s name).

I was also thinking that the inherent weakness of too easily removed links could be at least partially solved with acl’s and extended attributes. Of course that would also make upgrades more cumbersome, but that’s unavoidable: symlinks always cause problems for somebody. You can move the problem around, but it is always somewhere waiting to bite. That doesn’t necessarily mean that GoboLinux is a bad idea: symbolic links can solve a lot of problems too.

If you aren’t groaning and wondering how on earth I could be so stupid as to not see what an epically horrible idea all this is, you might want to poke through the GoboLinux Knowledge Base. If you are shaking your head and feeling sorry that I’ve lost all my marbles, trust me: I understand how you feel and why you feel that way. I’m not so enamored of this idea that I’m evangelizing it, but neither am I as against it as I thought I would be. I probably fall into some grey area between completely neutral and liking it, while at the same time reserving quantum rights to be mildly opposed at the same time. I guess I’m still highly ambivalent about GoboLinux.

*Originally published at APLawrence.com

Reddit | Furl

Bookmark WebProNews:

A.P. Lawrence provides SCO Unix and Linux consulting services http://www.pcunix.com

GoboLinux – A Breath of Fresh Air?
Comments Off
Top Rated White Papers and Resources

Comments are closed.