February 21, 2019

.comment: Not Forking But Branching - page 2

The Light at the End of the Tunnel Might Not Be a Train

  • February 21, 2001
  • By Dennis E. Powell

Over the last year or two we've seen what amounts to a forking among Linux distributions. There was a time when a binary for just about any distribution would run on just about any other distribution. There always have been, if memory serves, some structural differences in Slackware and Debian as compared to the other commercial distributions, but it was all well understood and under control. That's no longer the case. To pick an extreme example, an RPM for Red Hat 7 is not likely to bring satisfaction elsewhere. Though there has been a lot of noise about this, some of it from me, it does not represent a significant departure from Red Hat's philosophy of being first to market with things, some of which might not be quite ready.

There has been, too, considerable variation as to where particular programs are installed. This, too, has generated noise -- some of it from me -- because it makes things tougher, for some people tougher than for others. Those who get a distribution, install it, use it, and never deal with it otherwise until it's time to install a new version of the distribution have no reason to care where anything is installed. Those who upgrade binary packages do have reason to care: They'll need to know where everything is or make sure that the binaries are in RPMs for that version of that distribution, or try to redirect the RPM install, which may or may not work, because there may be other incompatibilities, too. Cross-distribution computability has gone out the window, and no matter which distribution you cheer for (as if it were a sports team), that's not a good thing. Those who are actively involved in development, or who like to try new stuff, are the ones most burdened, because they have to uninstall things put there by the distribution and reconfigure for them to go elsewhere, so that different versions can be kept track of, with the last working version as a backup. Throwing desktops into /usr, for instance, means that people who are compiling a new version are risking a lot, unnecessarily.

But these things have been embraced by distributors eager to have something, anything, that makes them different from every other Linux distribution. We have seen, too, the battle of the installation programs, the administration programs, and now the trend is toward the automatic online update programs (none of which involves the precious source code that the community is forever bleating about, by the way; the official desktop of the Free Software Foundation has lately been available to developers only in binary form, because its CVS tree won't build, so vast are the changes being made to core libraries by commercial concerns who have all but hijacked its development).

I have no reason to believe that what follows will happen, except that I think there's sense in it, but I have every reason to hope that it will, and Caldera's decision to eschew the retail channel fuels that hope.

What if Linux distributions were to specialize? We've already seen that the attempt to be all things to all people -- to produce incompatible distributions that otherwise do pretty much the same thing -- isn't working out all that well. What if distributions would instead target a segment of the market and go all out for that segment, and leave other segments to others?

Here's what I mean: Caldera is going after the enterprise. So, in their way, are companies such as IBM, to whom the Linux desktop means nothing. It's not as if there will be no competition there -- just that not everybody will be competing there.

Why not broaden this approach? There could, and I think should, be a few distributions aimed at the desktop user. These would have subtle but significant differences -- the one that has the very latest stuff, the one that puts stability and security above all else, one in the middle. These distributions would not install all kinds of servers that do nothing but open the machines to crackers; that's one significant difference over the current crop of one-size-fits-all distributions.

There are all sorts of other niches that would benefit from specially tailored versions of what we've come to regard as Linux (though the word strictly used is of course just the kernel).

The beauty of this is that there is space in the market for a lot more distributions to survive this way than they will in the current state of affairs in which everybody (except, now, Caldera) is chasing after everything. Moreover, the reasons to make one company's Linux base incompatible with other distributions would largely disappear.

Will this happen? Probably not. The urge to go for everything, to risk it all on one roll of the dice, is powerful, and reviled though it is, everyone wants to be the next Microsoft. The current Microsoft, though, is beginning to take notice and, wishing to continue to be the one and only Microsoft, is firing the opening shots in a war that's likely to get very ugly and very bloody. I think that the kind of specialization I've described is one way that Linux companies can emerge victorious. The Tower of Babel of distributions we have now is certainly not the recipe.

Most Popular LinuxPlanet Stories