Hacker News Reader

Comment by JoachimSchipper | Hacker News Reader

A while? You must have never read the UNIX-HATERS Handbook (http://simson.net/ref/ugh.pdf), which should be required reading for any unix fan, along with the Plan9 design documents (cat-v.org). In this case, "the X Windows disaster" chapter seems most appropriate. Feel free to ignore the VMS fanboys, though.

And yes, I am a dedicated unix (OpenBSD) user.

View on HN

As much as I love the Unix-Haters handbook, the more I learn about X the more I disagree with that chapter. X's memory consumption is totally reasonable for something that has to manage screen and window buffers (I doubt this proposed project would be any better). Xlib is retarded, but no one is forcing you to use it (esp. now that XCB is here). The old drawing protocol sucked, but XRender is very reasonable.

While I loved reading the Unix haters handbook it gets many things wrong. It gets a lot right too, but the point is it takes some critical thinking (and good knowledge of Unix) to tell which is which.

In that respect I found the chapter on X to be one of the weakest. They trot out the old "X named the client and server the wrong way around" chestnut for example...

Well, the naming is one thing; the architecture itself is worse. It's far more likely today for you to want to run applications on a server and be able to connect to and disconnect from them, like VNC or much better RDP, than it is for you to want all your applications running on the server to terminate when you disconnect your machine from the network.

There is nothing preventing any application to connect to another X display and push all its windows there, even if the original display session terminates. Emacs can sort of do that, for example, and there are some traces that something like that was intended even for GTK. But no application actually work in such conditions.

Isn't there xmove for this kind of stuff?

Yes, I completely agree: that's the "ignore the VMS fanboys" part.

It's a very good resource, but a rather uncritical collection of criticisms - some highlight important deficiencies, some highlight alternative implementations, some of it is out-of-date (most unices are quite stable nowadays), and some are just plain ignorant.

I recall the chapter on X being rather sensible, but I don't exactly love X either. Maybe that coloured my perception.

Someone should update the Hater's handbook. Many of the problems have been papered over, but few if any have been fixed. And I say this as someone who's lived inside Unix of one sort or another since '86.

Reply on news.ycombinator.com