Oh truly sorry for mentioning my favorite new feature in a comment rather than repeating the release notes, or trying to guess which one is the most useful to you.
GNOME Terminal introduced a new escape sequence that allows arbitrary text to become a hyperlink, pretty much like <a> tags on webpages. The same feature is also available in iTerm2's beta series as well as Tilix. Hope some of you will find it truly useful! As the first "official" tool to utilize this, coreutils-8.28 introduced "ls --hyperlinks". Filenames are linked to their corresponding "file://" URI, and ctrl+click conveniently opens them in their preferred graphical application.
I'm happy to announce that Gnome 2.10 includes the worst terminal emulator application ever seen, which obviously didn't survive more than one minute of testing. Just tap on Enter at your shell prompt and see it. Or try to view a manpage (with "less" as your pager). vte didn't have a maintainer for quite a long time, both Gnome 2.6 and 2.8 included vte 0.11.11 which suffered from plenty of major (but not critical) bugs (see their bugzilla) and many of them already had a patch either in their bugzilla, or in Fedora or some other distros. Some of these bugreports/patches were created by me, I've spent a couple of days making vte suck less, but as I'm not a Gnome developer and don't have CVS access, all I can do is put every piece of information I know into their bugzilla. On Mar 3 (less than a week ago) some random patches were committed into cvs, which made the very basic features of vte (e.g. being able to scroll inside the terminal) go wrong. Also the new maintainer complained that some of my patches no longer apply and shall I please port them to the new codebase. Okay, "patch -p1 blabla" isn't able to apply them anymore but it would take about two minutes to apply them manually, especially if he read the comments where I described what and why I did. (Now I'm talking about bug 164153.) So he applied untested and faulty patches, didn't apply some good patches just because they didn't apply cleanly with the "patch" utility, and then released vte-0.11.12 which is so fscking broken that I can't understand how someone could work with it for more than twenty seconds. Is this really the way to go? Commit tons fundamental non-trivial patches just five or six days before release so that there remains no time to test them? And then don't even try to test them 'cause I'm pretty sure that everyone would notice within a minute that the current version is unusable? Aren't there some development policies that would forbid such kind of change-everything-right-before-major-release commits?
check out the source of "pine" or "wu-ftpd"! you'll realize that everybody in the world writes beautiful code except people at the washington university.
If K numbers are chosen from a set of N elements (in the above example N=365 and K=28), and p is the probability of collision, and p is not close to 1, then approximately this holds:
-ln(1-p)=(K(K-1))/(2N).
If p is very small (that's what we want in the pad method) then this gets even simplier:
p=(K^2)/(2N).
In the pad algorythm N=2^64, so if we say let p=1/10^6=1/2^20 then we get K=2^22.5, so at most 6 million pads can safely exist. Is it enough? I don't think so. (Think of CDDB which also sucks.) I suggest that the filename contain at least 128 bits of data instead of only 64. For example the filename could contain the MD5sum of the whole pad.
Any malicious user can generate pads which have the same filename as an already existing pad. The filename should contain a hash (eg. md5) of the whole pad file and the mirroring utilities should always check this.
Slashdot Mirror
Oh truly sorry for mentioning my favorite new feature in a comment rather than repeating the release notes, or trying to guess which one is the most useful to you.
Wrong escape sequence for Shift-F1? Yup that was fixed 3.5 years ago.
GNOME Terminal introduced a new escape sequence that allows arbitrary text to become a hyperlink, pretty much like <a> tags on webpages. The same feature is also available in iTerm2's beta series as well as Tilix. Hope some of you will find it truly useful!
As the first "official" tool to utilize this, coreutils-8.28 introduced "ls --hyperlinks". Filenames are linked to their corresponding "file://" URI, and ctrl+click conveniently opens them in their preferred graphical application.
I'm happy to announce that Gnome 2.10 includes the worst terminal emulator application ever seen, which obviously didn't survive more than one minute of testing. Just tap on Enter at your shell prompt and see it. Or try to view a manpage (with "less" as your pager).
vte didn't have a maintainer for quite a long time, both Gnome 2.6 and 2.8 included vte 0.11.11 which suffered from plenty of major (but not critical) bugs (see their bugzilla) and many of them already had a patch either in their bugzilla, or in Fedora or some other distros. Some of these bugreports/patches were created by me, I've spent a couple of days making vte suck less, but as I'm not a Gnome developer and don't have CVS access, all I can do is put every piece of information I know into their bugzilla. On Mar 3 (less than a week ago) some random patches were committed into cvs, which made the very basic features of vte (e.g. being able to scroll inside the terminal) go wrong. Also the new maintainer complained that some of my patches no longer apply and shall I please port them to the new codebase. Okay, "patch -p1 blabla" isn't able to apply them anymore but it would take about two minutes to apply them manually, especially if he read the comments where I described what and why I did. (Now I'm talking about bug 164153.)
So he applied untested and faulty patches, didn't apply some good patches just because they didn't apply cleanly with the "patch" utility, and then released vte-0.11.12 which is so fscking broken that I can't understand how someone could work with it for more than twenty seconds.
Is this really the way to go? Commit tons fundamental non-trivial patches just five or six days before release so that there remains no time to test them? And then don't even try to test them 'cause I'm pretty sure that everyone would notice within a minute that the current version is unusable? Aren't there some development policies that would forbid such kind of change-everything-right-before-major-release commits?
check out the source of "pine" or "wu-ftpd"! you'll realize that everybody in the world writes beautiful code except people at the washington university.
If p is very small (that's what we want in the pad method) then this gets even simplier:
In the pad algorythm N=2^64, so if we say let p=1/10^6=1/2^20 then we get K=2^22.5, so at most 6 million pads can safely exist. Is it enough? I don't think so. (Think of CDDB which also sucks.) I suggest that the filename contain at least 128 bits of data instead of only 64. For example the filename could contain the MD5sum of the whole pad.
Any malicious user can generate pads which have the same filename as an already existing pad. The filename should contain a hash (eg. md5) of the whole pad file and the mirroring utilities should always check this.