In compiling my report about Linux for uni, I have merely skirted around the topic of LSB. I notice, for example, that there are only a few distros that have LSB compliance.
I guess that the LSB is a good idea (in theory), since Linux has gone the way UNIX did by diversifying (perhaps too much)... Does anyone have any thoughts on this? It seems, also, that compliance means that RPM needs to be used? (That's insane IMHO, but I haven't tried APT-RPM).
Any thoughts oh wise ones? ;)
Ricardo Campos ricardo@corez23.com wrote:
I guess that the LSB is a good idea (in theory), since Linux has gone the way UNIX did by diversifying (perhaps too much)... Does anyone have any thoughts on this?
Yes. Probably the problem has to be solved by embracing our similarities instead of trying to hide the difference. If I knew how best to do that, I'd probably be rich by now ;-)
It seems, also, that compliance means that RPM needs to be used?
I think this is bogus: RPMs need to be installable, and SysV scripts too, but they can be done through compatibility layers (in the style of Alien?) rather than by using RPM as the system packaging format.
However, I do have reservations about LSB picking the vendor-tied RPM format. The format has had multiple incompatible versions which have been unleashed on an unsuspecting world (because to announce the imcompatibilities too loudly would hurt sales, I guess) and caused pain as the world made the transition. I think it would have been better to pick a non-vendor-tied binary package format without the distribution-specific dependency handling, such as Encap, perhaps.
Any thoughts oh wise ones? ;)
Brain dump above. Disagree at will.
Vendor specific installers? <shudder> Shades of MSIE.
The problem with any standard like LSB is that everyone has their own belief as to what its essential to include in the standard, with the result that the final document is a bloated list of everyones' favorite widget.
Trimming such a document to the point where it actually becomes useful is a task for Solomon! The golden rule should be observed ....
"Have as few rules as possible. Then there are fewer rules to be broken. If you can get away with no rules at all, you achieve utopia ... a crime-free state."
TD
On Sun, 2003-01-12 at 23:04, MJ Ray wrote:
Ricardo Campos ricardo@corez23.com wrote:
I guess that the LSB is a good idea (in theory), since Linux has gone the way UNIX did by diversifying (perhaps too much)... Does anyone have any thoughts on this?
Yes. Probably the problem has to be solved by embracing our similarities instead of trying to hide the difference. If I knew how best to do that, I'd probably be rich by now ;-)
It seems, also, that compliance means that RPM needs to be used?
I think this is bogus: RPMs need to be installable, and SysV scripts too, but they can be done through compatibility layers (in the style of Alien?) rather than by using RPM as the system packaging format.
However, I do have reservations about LSB picking the vendor-tied RPM format. The format has had multiple incompatible versions which have been unleashed on an unsuspecting world (because to announce the imcompatibilities too loudly would hurt sales, I guess) and caused pain as the world made the transition. I think it would have been better to pick a non-vendor-tied binary package format without the distribution-specific dependency handling, such as Encap, perhaps.
Any thoughts oh wise ones? ;)
Brain dump above. Disagree at will.
-- MJR http://mjr.towers.org.uk/ IM: slef@jabber.at This is my home web site. This for Jabber Messaging.
How's my writing? Let me know via any of my contact details.
main@lists.alug.org.uk http://www.alug.org.uk/ http://lists.alug.org.uk/mailman/listinfo/main Unsubscribe? See message headers or the web site above!
Tony Dietrich td@transoft.demon.co.uk wrote:
The problem with any standard like LSB is that everyone has their own belief as to what its essential to include in the standard, with the result that the final document is a bloated list of everyones' favorite widget.
This need not happen, though. In some cases, the standards group have decided to make the first standard a maximal subset of common features, as far as possible, and then layer on top of that over time, as consensus is reached. That doesn't seem to have happened with LSB for some reason, but it has happened with some popular recent programming languages. I know that Scheme is developing this way, with the IEEE and RnRS specifications being built upon by the SRFI process. I think Python's PEPs and Ruby's RERs work in a similar way, but those languages are specified by one implementation.
On Fri, Jan 10, 2003 at 07:02:04PM -0000, Ricardo Campos typed the following...
In compiling my report about Linux for uni, I have merely skirted around the topic of LSB. I notice, for example, that there are only a few distros that have LSB compliance.
I guess that the LSB is a good idea (in theory), since Linux has gone the way UNIX did by diversifying (perhaps too much)... Does anyone have any thoughts on this? It seems, also, that compliance means that RPM needs to be used? (That's insane IMHO, but I haven't tried APT-RPM).
Any thoughts oh wise ones? ;)
...and that's all Ricardo Campos wrote I'm afraid
Aaaarrrrgggghhhh, I replied to this and forgot to make sure that it was copied to the list as well - I wish there could be some standardisation here! Some lists you just reply, some you need group reply and a very few actually have the headers for a special list reply - I've read both the arguments on reply-to munging, and although I'm not completely decided I think I prefere to be able to reply to the list rather than group reply (most of the time you reply to the list, and if a thread gets too long you end up with a really long copy list if its not trimmed, and I also get fed up with getting two copies of replies to my own posts). Anyway, I've probably started a whole new sub-thread here - my real reply is below...
-
I like the idea of standardising the directory structures so that you know where to find things when you switch distributions - that certainly shouldn't be the way a distro different. The package management side of things I'm not so sure about, but I'm not 100% sure of my motives here. I can see the attraction of having a common package format that will install across all distributions, but...
I can see why RPM was chosen as the standard on the practical side of things (i.e. its used on a wide range of distros), but on the technical side I really couldn't bring myself to go back to using it having use the Debian system. I don't really care if RPM is available, so long as it doesn't take over and can remain an irrelevance to me :-)
Personally I think that with things like package management there is room for a little choice, and the LSB could devine a couple of alternatives. We can manage with different standard for other applications, video and audio formats, etc., so as long as there aren't too many I would say that choice was a good thing - there's a balance to be had between the 'one size fits all' Windows way, and having so many choices.
Just my 2p's worth anyway.
Summary:
Paul Tansom paul@whaletales.co.uk wrote:
[...] I wish there could be some standardisation here! Some lists you just reply,
That is broken as it removes the "reply off-list" possibility from some email clients and publishes when people may not have made a conscious decision to publish. Find me an email client without "reply to all" and I'll show you an antique in need of an upgrade.
some you need group reply
Always use "reply to all" and it should never fail to reach the list.
and a very few actually have the headers for a special list reply
RFC2919 or similar and we support them, I believe.
So, we do the most that we can do to make sure that replys for the list go to the list without tripping up unsuspecting readers in what can be a very embarrassing way. I think that's fair and I'll try to obstruct anyone trying to screw it up now ;-)
For example, have you ever emailed 3000 people around the world working in the same field as you with a "hi, haven't heard from you for ages, how's the new house?" email? I have. (Yes, I know *NOW* to check where Reply-To points... can't put the toothpaste back in the tube.)
On Mon, Jan 13, 2003 at 11:19:26PM +0000, MJ Ray typed the following...
Summary:
Paul Tansom paul@whaletales.co.uk wrote:
[...] I wish there could be some standardisation here! Some lists you just reply,
That is broken as it removes the "reply off-list" possibility from some email clients and publishes when people may not have made a conscious decision to publish. Find me an email client without "reply to all" and I'll show you an antique in need of an upgrade.
Not quite sure how it removed the reply off-list option, but that may just be because I've not been playing with email clients for a bit. I always get the option of using the from or reply-to address. That said I have a feeling that a number of clients must actually ignore the from and reply-to headers as I've sent a good number of emails out recently have have been sent back to the real sending address instead of either the from or reply to I've configured in the email - either that or my email client (Mutt) is doing something odd.
some you need group reply
Always use "reply to all" and it should never fail to reach the list.
Its all right, I usually do, I've just been doing a lot of mail recently and have also made an annoying number of slip ups where I've just replied.
and a very few actually have the headers for a special list reply
RFC2919 or similar and we support them, I believe.
No, I don't believe you do, unless the Mutt List Reply option uses an alternative method. Using L to reply to list gives an error stating no list found.
So, we do the most that we can do to make sure that replys for the list go to the list without tripping up unsuspecting readers in what can be a very embarrassing way. I think that's fair and I'll try to obstruct anyone trying to screw it up now ;-)
For example, have you ever emailed 3000 people around the world working in the same field as you with a "hi, haven't heard from you for ages, how's the new house?" email? I have. (Yes, I know *NOW* to check where Reply-To points... can't put the toothpaste back in the tube.)
...and that's all MJ Ray wrote I'm afraid
No, I've not done that, I usually end up sending a replying to too few people, mainly with lists where I forget to group reply. I also have had a few emails I've sent out for long threads where I've gone in and deleted not only myself (because I use multiple addresses), but another 10-15 addresses from the cc list.
Anyway, for me I think this sub-thread is finished. Its really not a big enough deal for me to worry either way (not sure why I got the urge to mention it really). I can see both sides and have not yet decided what format to use for a couple of lists I'm setting up at the moment. I would, out of curiosities sake, be interested in figures/estimates on the amount of extra traffic caused by copying recipients who will also receive the mail via the list vs extra traffic caused by inadvertantly replying to too many people - I very much doubt there is anything though.
Anyway, nuff said :-)
On Tue, Jan 14, 2003 at 01:26:55PM +0000, Paul Tansom wrote:
and a very few actually have the headers for a special list reply
RFC2919 or similar and we support them, I believe.
No, I don't believe you do, unless the Mutt List Reply option uses an alternative method. Using L to reply to list gives an error stating no list found.
you havn't told your mutt which lists are mailing lists, my muttrc has two lines
lists main@lists.alug.org.uk subscribe main@lists.alug.org.uk
which means that the L will work, you want to read this for more info on it. http://www.mutt.org/doc/manual/manual-4.html#ss4.8 it also explains how the reply-to etc. are handled by mutt.
Adam
Paul Tansom paul@whaletales.co.uk wrote: [Reply-to MUNG]
Not quite sure how it removed the reply off-list option, but that may just be because I've not been playing with email clients for a bit. [...]
OK, normally email clients have two options: "reply" which uses the Reply-To line if set, or the From line otherwise; "reply to all" which adds the To and Cc contents to that. Because some mailing lists started to reset Reply-To to point to the list, some email clients (including mutt) ask you whether to use Reply-To or From if both seem valid. However, if the From address wasn't the address where off-list replies should go, removing the original Reply-To removes the off-list reply possibility...
This horrible situation has made Reply-To basically useless in many situations and a range of ugly non-standard hacks and kludges have grown up, including Mail-Followup-To.
and a very few actually have the headers for a special list reply
RFC2919 or similar and we support them, I believe.
No, I don't believe you do, unless the Mutt List Reply option uses an alternative method. Using L to reply to list gives an error stating no list found.
Yes, we do. I just checked and the List-Id, List-Post and other headers are present on ALUG mailing lists.
As of October 2002, when I last checked, mutt did not support these headers and required you to add list email addresses to your configuration file in order for the "list reply" command to work. I'm sure they'd love a patch.
[...]
..and that's all MJ Ray wrote I'm afraid
http://learn.to/attribute while I'm being prescriptive.
No, I've not done that, I usually end up sending a replying to too few people, mainly with lists where I forget to group reply. [...]
Well, the damage from that is limited to normally having to forward something from the sent mail folder, which is much less severe, IMO.
Anyway, for me I think this sub-thread is finished. [...]
OK, np.
I would, out of curiosities sake, be interested in figures/estimates on the amount of extra traffic caused [...]
Most of the list hosts which take advertising force reply-to to point to the list, which I think is probably an indication that it is significant ;-)
On Tue, Jan 14, 2003 at 03:41:09PM +0000, MJ Ray typed the following...
http://learn.to/attribute while I'm being prescriptive.
...and that's all MJ Ray wrote I'm afraid
OK, it took me a while to work this one out, and I've decided that, apart for being a bit verbose the main difference is the lack of the email address, so given that I've deliberately chosen not to include that in mine I thought I'd post why - oh no more controversy ;-)
Basically I avoid quoting emails (other than my own) to lists where possible. To be honest the main source of issues with this is the SmoothWall list (less said the better), but it has to be said that since the SmoothWall list was made available as a web archive (now closed again), my spam quota has rocketed, and become more unpleasant. I now get some really distasteful porn spam, and this is undoubtedly harvested (apart from the obvious fact that I don't frequent sites of that sort) because it comes to paul@smooth..., community@smooth... as well as my personal paul@whale... and paul@apta... addresses. 99% of the time there is no way to unsubscribe even though some give details of how to (the domains usually don't exist or are not active).
Took a quick look at the ALUG archives and I see that this is not considered a problem here (no hiding of email addresses at all), so I suspect that I know your views here ;-)
On Wednesday 15 January 2003 12:57, Paul Tansom wrote:
Basically I avoid quoting emails (other than my own) to lists where possible. To be honest the main source of issues with this is the SmoothWall list (less said the better), but it has to be said that since the SmoothWall list was made available as a web archive (now closed again), my spam quota has rocketed, and become more unpleasant
I see your point, ALUG and personal email's are about the only place I use my *real* email address. ALUG because after looking at how the list archives are presented I felt fairly safe.
The amount of spam I received did suddenly skyrocket late last year, however the interesting thing is that an additional btinternet account I have but have never sent from has also started receiving unsolicited mail ! Figuring that someone must have figured out a way to harvest BT's registration system for valid addresses. I contacted BT but have never had a sensible reply to confirm or deny this.
On a side note-
Now I use SpamAssassin, and I must say that as a completely maintance free spam killer it is fantastic ! I get about a 80% hit rate on unsolicited mail, but the most important thing is that I have NEVER had a false positive.
On Wed, Jan 15, 2003 at 07:54:46PM +0000, Wayne Stallwood wrote:
Now I use SpamAssassin, and I must say that as a completely maintance free spam killer it is fantastic ! I get about a 80% hit rate on unsolicited mail, but the most important thing is that I have NEVER had a false positive.
I find a combination of rejecting mail with non English content types (Content-Type:.*charset="?(ks_c_5601-1987|gb2312|euc-kr)"?) and using bogofilter helps a lot; I used to use the Content-Type stuff and SpamAssassin, but found it missed a fair bit. While I'm still in the process of training bogofilter the main thing it gets as false positives are abuse reports with spams attached...
J.
Jonathan McDowell noodles@earth.li wrote:
I find a combination of rejecting mail with non English content types (Content-Type:.*charset="?(ks_c_5601-1987|gb2312|euc-kr)"?) and using bogofilter helps a lot; [...]
If you add the DCC filters http://www.rhyolite.com/anti-spam/dcc/ in between these two checks, with a suitable whitelist, you should get something similar to my current set-up, probably with a similarly high success rate.
Paul Tansom paul@whaletales.co.uk skribis:
[...] apart for being a bit verbose [...]
That was it ;-)
the SmoothWall list was made available as a web archive (now closed again), my spam quota has rocketed, and become more unpleasant. [...]
Well, basically, ALUG archives include email addresses both as a form of attribution and as a way for people to contact you about your message. Sadly, some unscrupulous individuals will take email addresses from whereever they can get them, including such list archives but also newsgroups, other web pages, any email records or subscriber lists, nameserver lookups of certain ISPs, domain registrations, etc. About the only way to avoid spam completely is to avoid the Internet completely.
For the rest of us, spam filters are the order of the day. Some people will prefer regular expressions, some spamassassin, some a Bayesian filter, some the DCC project, some DNS blacklists, and some a combination. I hope you find whatever works for you.
On Thu, Jan 16, 2003 at 10:13:49AM +0000, MJ Ray wrote:
Jonathan McDowell noodles@earth.li wrote:
I find a combination of rejecting mail with non English content types (Content-Type:.*charset="?(ks_c_5601-1987|gb2312|euc-kr)"?) and using bogofilter helps a lot; [...]
If you add the DCC filters http://www.rhyolite.com/anti-spam/dcc/ in between these two checks, with a suitable whitelist, you should get something similar to my current set-up, probably with a similarly high success rate.
I'm wary of this sort of thing; I've seen reports from people that many legitimate mails are tagged by people using Razor. Does DCC suffer from this?
J.
Jonathan McDowell noodles@earth.li wrote:
I'm wary of this sort of thing; I've seen reports from people that many legitimate mails are tagged by people using Razor. Does DCC suffer from this?
DCC is a different filter -- it aims to detect and tag bulk email (even with some variations such as the mailmerge name substitutions done by spammers) rather than detect spam. You should either filter your mailing lists before feeding to DCC (my solution) *or* whitelist them in some way.
I've had one person get tagged and bucketed for sending me more copies of an email than my threshold number. Not too severe, as I empty the bucket daily (ie look, select all, delete most days).
-
Adam Bower -
Jonathan McDowell -
MJ Ray -
MJ Ray -
Paul Tansom -
Ricardo Campos -
Tony Dietrich -
Wayne Stallwood