On Sat, 19 Jul 2008, Alan Bartlett wrote: >> Have you replied to Russ Herrold's comment? Perhaps I missed it. >> >> No, but others did and I agree with their comments. > I think it would have been polite of you to have responded to Russ' > comments. (Speaking as an Englishman.) Polite to one side, as proponent, he needs to so speak; I had deemed the proposal dead -- more, infra. > Basically I think that pointers to the outside are not that good >> because they not necessarily contain CentOS specific information. and frankly, if SRPM building is being described in a non-portable fashion, or in a fashion which of a local distribution 'dialect' it is LESS good than the portable one. One might ask why: One thing which the Fedora folks do which destroys portability is to add manual non-essential BR's or versioned BR's when NOT required, to _force_ a person to either follow their development path, or take the time to identify that it is a false one, and cut it out. A second related one is to manually insert false manual Requires, which affect install time. A third is a 'the latest is the greatest' mentality, and always developing with the lastest bleeding edge tools, which destroy prior version portability. As such, getting say, spacewalk working on CentOS 3 will be a major undertaking. Or the fine work which Mike DeHann is doing at the ET Labs -- he used a non-compatible w RHEL 4 Python module, and so unless one is running 5 series, one cannot use it. Really and truly, CentOS (as a community re-statement of the upstream's comercial product) is a long lived distribution, and content we issue (including documentation) should have a seven year supported livespan -- This means NOT particularizing it to a specific CentOS variant as much as possible -- the x11 to xorg cutover comes to mind -- it causes me heartburn making sure xpdf and graphviz still back-build over time on all three platforms. Really, in RPM space, if there is NOT A NEED for it. > The main link that Akemi and I currently have in Kernel > Sources and Custom Kernel is to the (maintained) CentOS > specific Owl River page ( > http://www.owlriver.com/tips/non-root/). A CentOS channel member pointed out a variance from an RFC form, and I fixed on the spot, adding attribution, just this week, in the 'tips' hierarchy. > The other link, in the Custom Kernel article, > (http://howtoforge.com/kernel_compilation_centos/) is > mentioned to strongly dissuade its use. HowToForge has LOTS of bad and immutable content -- the one about how to 'harden' CentOS is a horror, with lots of unmaintained 'compiled from tarballs, and not under package managercontrol.' The Bastille project was another really bad example of fighting the packaging system. >> Anyway, building RPMs is something I learned the hard way >> because there is no one-stop easily-understandable >> comprehensive but howto-like document, The GuruLabs people will be very surprised. They regualrly teach a rather good course in just that in the curriculum which I have obtained their express consent to copy and place for public access, back in the 2000 to 2007 period when I served as the Editor of the RPM website. You are simply wrong as to the resources out there. Ed Bailey's book, and Eric F-J's also completely cover the matter and each are freely available on line, and linked to already through the CentOS wiki. >> and that is what I was trying to address. But it's your >> decision as to if the idea of creating this on CentOS wiki >> has merits or not. I have not heard you express it, and frankly that silently you agree with some of the posts in the thread requires me to 'read your mind' to know your position as proponent. I decline to so speculate. I have trimmed a bit, but feel free to write content, and point me at it publicly or privately, and I'll review and comment. I'm off to OLS, driving up with JBJ this week, and if I get stuck, I'm sure the long time principal developer of RPM can help me through any confusion. Red Hat's current maintainer seems not to be shy about 'cribbing' fixes from JBJ's RPM5 project, and I know that both JBJ and I read EVERY new RPM related bug that crosses up the upstream's tracker. > Russ, Ralph and Akemi - do you have any comments, please? As in my prior email written before I saw this one, and as above. -- Russ herrold
