like any project, the best was to start is still thinking about dividing the original problem into small chunks. setup some rough timeline and find the correct person for each task to take over.
one big point would be all kind of administration and distribution of settings. maybe, LDAP will be the account storage of choice, but maybe, there are some other reasons against it (e.g. the need for a possibility to bulk change user properties fast and easily).
another big point is mail storage, affecting all kind of clustering you think of. huge and redundant NFS boxes are open for changes, but other storage solutions may be faster,....)
or your network design? your smtp hosts will receive tons of mail every day, should they be the same as the ones doing virus/spam checks and bouncing of non-existant accounts? it's not very useful to check inbound mails you dont's have a "local" recipient for. atm, at work, i use a setup, where the first 2 smtp hosts (load balanced, but could be via dns too) just check if the account ldap if the account exists + some minor checks, to relay internally to the next step (this way, there are free ressources for larger worm and virus waves). step 2 is virus scanning and spam detection (just adding headers, bigger machines, the cpu intensive part). finally, step 3 is the correct delivery system that stores the message itself in the user's mailbox. at the same time, networking issues including firewall, internal interfaces and administrative access should be part of this considerations.
the easy setup combination point: after some research, the best for the last delivery and user access i found was qmail-ldap and courier. in your case, that depends on your requirement specification. comparatively easy questions like the mailbox format of choice should also be solved by this group.
security has to be an issue for all parts. how are you planning to secure your outer smtp? are there requirements that people from other networks (e.g. at home) use your smtp for sending mails (not using webmail, that'd be easier)?
monitoring & backups - 99.9% is a hard task, find the correct way of monitoring parts where you expect problems.
and in the end, the group where you will mostly spend your time: side effects of unclear specification.
Re:it's simple, but...
on
Just a Phone?
·
· Score: 2, Interesting
my gramdma has a "rufhilfe" button to wear like a watch that works within a 200m radius at home (connected via a base station, connected to pots like a modem). our attempts to give her a possibility of calling help when hiking or on mushroom foray failed because of to small buttons on cell phones.
unless a phone does not represent the habits she already has and is not really willing to change (why should she?), a menu more simple would not really help her. she never looks at the display if someone calls her, she's not interested in receiving or sending SMS. a perfect cell phone for elder people shold have extra quick dialing buttons, a small display (it isn't used anyway) and access to a configuration menu via some key combination they don't happen to fall into, after the grandchild or the elderly care nurse had configured it once for their needs.
it's not yet perfect, but i'm glad to see improvements.
pascal itself is outdated, but it still is a very clear language where coding is just great fun.
the idea of lazarus (write once, compile on different platforms) is very nice, the linux ide still had some nasty bugs but provides the most important features AND it was more stable than the one from kylix.
oracle is interested in linux, but not the way one may be used to from bigger players. they neither are ibm nor novell.
some years ago, i downloaded oracle beta for linux (i didn't want to be blamed for spending the money for an aix development system) and the next day, i had a phone call by a local oracle office to tell them about my experiences including free training courses (i didn't intend). it was not one of those alibi questions "how's my product *extra-white-smile*? wanna buy cheap rdbms now?", but a serious attempt to keep in touch with people that really use and evaluate it. direct questions by technicians and not the usual marketing blah. i really was surprised.
oracle has some not too big focus on linux, when it's redhat or sles and will hardly ever satisfy the kind of linux lover running slackware, debian, gentoo or his own flavour. they have never seen the benefits of linux in common because they never needed to.
for oracle, linux is unstable and inscalable if larry sais so until he sais anything different. if larry hasn't said anything yet, don't really care about their statement in the meanwhile;)
in some countries, you get a drivers license (for anything you want) by simply setting the creation time of a file on a floppy disk. and who had not root access at school?
it is cheating and it has to be punished, but i think this pupil just took the kirk solution for kobayashi maru test and it should not be too hard. the others that would have bought the test results are the real problem (and if they would not exist, it's questionable if he would have done it).
the thinkpad i got provided by some company doesn't even have windows keys, but two nice grey "switch terminal" buttons - just perfect for linux.
it still was in the original packing and had no windows installed, only a small "designed for windows XP" sticker - that i kept for amusement.
and after all notebooks of other vendors (HP, compaq, acer) where i chose to run on linux, i'm really satisfied with my thinkpad.
of course, i had to tweak lots of things for myself, but very often i came across ibm e-mail addresses within the sources i had to compile.
maybe they are not yet self confident enough to promote their linux support on their own (i mean, not any more their own but using their brand name) notebooks, but i'm already pleased by the few things they provide - it's at a good start at least.
as i started to love java, i was forced to learn PHP - and i did not like it, as any other script language, i considered it to be just another programmer newbee easy-to-learn and automate-anything-for-me language.
i argued against the decision of implementing a huge-scale web site in php, but i lost. so i had to deal with it and read the php & zend engine sources (just a habit when dealing with an interpreter) and found out what it obviously is: easy to use glue code, where you have to know more about type casting, cost of algorithms and external functions than in java, if you want your huge project to perform. even the question where to hold what kind of resource (persistancy considerations), where to implement the abstractions gets a hard nut each time. but after all, you get a very fast, perfectly scalable (just another box) application, no more dealing with fail-over stragegies - and if you don't have time to implement some special, hardly ever used feature, you get cheap external resources.
of all arguments i had against php, the strongest i still have is, that there are lots of developers that call themselves to know php and all you may expect them to code is an autoglobal using guestbook.
java is still great - and for some middle-scale applications, i still consider using it (besides all the other project dependent reasons for the language of choice) - but if it get's really big, i wouldn't use it any more since i experienced this kind of php usage. (also proven by some j2ee to php migrations we did over the last 2 years)
you just talk about one of the big advantages of PHP - because it _does_ use this amount of external libraries. (and a pain in the ass for thee ones that have to update their special-purpose compilation right now)
the wrappers are just needed for the zend engine to know how to map php-function parameters to the exports of any external lib (plus some extra things like constants or not too big abstractions for convenience and error handling).
your examples GD and MySQL happen to address two exceptions for good reasons. the history of GD deveolpment, bug-fixing and reachability of the maintainer just proposed this kind of bundle (e.g. the current gdlib has some build problems on solaris again) take a look in the config.m4 in/ext/gd of your php sources to see the crossing of bundled vs. external lib. for the bundled MySQL, there are some other reasons, leading from a tight cooperation of MySQL AB, Zend and some core developers.
but "php does not link enough external libs" is absolutely wrong.
the first announcement was on wednesday by stefan esser on bugtraq and full-disclosure. additional info came up later on, PHP 4.3.10 changelist and the according changes in CVS can be seen by everyone since then.
i do not think that any serious system administraton team gets to know this vulnerabilities via slashdot for the first time, the problem is, that compilation of all dependencies and deployment to all systems is a hard job to do from wednesday night to a reasonable end just before weekend.... in a week where lots of companies have their *drinks all inclusive* xmas events and the usual pre-vacation problems occur.
the w3schoole table is rather incomplete -
consider using php.net operators manual instead - there are more of them (but far away from maxing out ascii like perl 6 is about to do)
[i]"can i get one more of your sweets? i need 'em for this special move in street fighter"[/i]
greets from the inventor of the "doom 2 walkthrough chewing gum";)
the ibm & ms divorce took place while there were 2 versions of os/2 in development v2 & v3 - and, primitively spoken (omitting some nasty war of roses), ms got parts of the v3 code base. to get a real os from it, they needed the cited programmers - and got vms developers.
this could have been a reason for windows NT starting with version 3.0 (and it's as likely as any other theory) and the mixture between ibm & vms way of problem-solution. a microsoft "own way" did not exist that time.
in the time we are talking about, microsoft was not that big knowledge center of development - but an interesting franchise strategy in the beginning of getting megalomaniac.
my connection is fast enough - but my gambing windows is always out of date (i don't use it more often than once a week on lan games) - so i download updates via linux and install from a fat32 partition before connecting a possibly risky windows to any network. as it's a notebook and a reinstalllation may be anywhere i have the time to play windows (train, plane, boring multihour kickoff meetings), it's better than a cd i'm sure to forget anywhere.
have you ever tried windows update under IE/wine? it's useless, but works;)
i think this argue was exaggerated. i'm used to synchronised movies & series and heard more than 5 different sean connery voices while edward norton is the same speaker as gauron in lord of the rings. it's not the sound of simpsons that makes me a fan , i'd really be disappointed if they ran out of yellow color;-)
i'm used to the old installer and i learned to stop worrying and love ncurses frontends. despite of some naming conventions, i can't see much difference between freebsd, solaris and old debian installers.
ad knoppix:
once, i tried to do a knoppix auto-install (make own knoppix cd that installs itself on boot, starts networking (dhcp) and runs sshd for easier server installation (i forgot the special supersonic ibm kvm cable and was too curious about the new toy...) - it was unlikely to work, but in the end, there only was one single annoying point: knoppix didn't install the initrd image it uses itself from cd.
if a system runs from a knoppix cd, this does not mean, that it still runs after a knoppix-install, espacially with scsi inside...
each EU member has to merge the new restrictions (less cadmium, less lead, forced recycling,...) into their national law until 2005. it really starts on 1.1.2006 and concerns reduction of lead, mercury, cadmium & other heavy metal.
in summer 2001, philips semiconductors, infineon & some other committed to prduce nearly unleaded chips.
php, apache & tools i link against oracle - things i always compile myself in order to avoid problems.
if you work in a heterogenous environment, this is one of the possibilities of keeping things even a lil' bit similar - and as a bonus, it's more stable.
very common things like the most common gnu tools, i get from sunfreeware, debian, as ports or via cygwin - and i'm used to a lack of goodies (ahh- tar knows "z" here?).
what really saves me a lot of time is a collection of shell scripts containing the configure strings and some written dependency trees (compiling the next missing lib on next term gets boring after a while - and too much coffee a day keeps the girlfriend away)
Slashdot Mirror
divide et impera
like any project, the best was to start is still thinking about dividing the original problem into small chunks. setup some rough timeline and find the correct person for each task to take over.
one big point would be all kind of administration and distribution of settings. maybe, LDAP will be the account storage of choice, but maybe, there are some other reasons against it (e.g. the need for a possibility to bulk change user properties fast and easily).
another big point is mail storage, affecting all kind of clustering you think of. huge and redundant NFS boxes are open for changes, but other storage solutions may be faster,....)
or your network design? your smtp hosts will receive tons of mail every day, should they be the same as the ones doing virus/spam checks and bouncing of non-existant accounts? it's not very useful to check inbound mails you dont's have a "local" recipient for. atm, at work, i use a setup, where the first 2 smtp hosts (load balanced, but could be via dns too) just check if the account ldap if the account exists + some minor checks, to relay internally to the next step (this way, there are free ressources for larger worm and virus waves). step 2 is virus scanning and spam detection (just adding headers, bigger machines, the cpu intensive part). finally, step 3 is the correct delivery system that stores the message itself in the user's mailbox.
at the same time, networking issues including firewall, internal interfaces and administrative access should be part of this considerations.
the easy setup combination point: after some research, the best for the last delivery and user access i found was qmail-ldap and courier. in your case, that depends on your requirement specification. comparatively easy questions like the mailbox format of choice should also be solved by this group.
security has to be an issue for all parts. how are you planning to secure your outer smtp? are there requirements that people from other networks (e.g. at home) use your smtp for sending mails (not using webmail, that'd be easier)?
monitoring & backups - 99.9% is a hard task, find the correct way of monitoring parts where you expect problems.
and in the end, the group where you will mostly spend your time: side effects of unclear specification.
my gramdma has a "rufhilfe" button to wear like a watch that works within a 200m radius at home (connected via a base station, connected to pots like a modem). our attempts to give her a possibility of calling help when hiking or on mushroom foray failed because of to small buttons on cell phones.
unless a phone does not represent the habits she already has and is not really willing to change (why should she?), a menu more simple would not really help her. she never looks at the display if someone calls her, she's not interested in receiving or sending SMS. a perfect cell phone for elder people shold have extra quick dialing buttons, a small display (it isn't used anyway) and access to a configuration menu via some key combination they don't happen to fall into, after the grandchild or the elderly care nurse had configured it once for their needs.
it's not yet perfect, but i'm glad to see improvements.
pascal itself is outdated, but it still is a very clear language where coding is just great fun.
the idea of lazarus (write once, compile on different platforms) is very nice, the linux ide still had some nasty bugs but provides the most important features AND it was more stable than the one from kylix.
still intersting and in some kind very funny the tanenbaum vs. torvalds debates about microkernel vs. monolithic architecture
oracle is interested in linux, but not the way one may be used to from bigger players. they neither are ibm nor novell.
;)
some years ago, i downloaded oracle beta for linux (i didn't want to be blamed for spending the money for an aix development system) and the next day, i had a phone call by a local oracle office to tell them about my experiences including free training courses (i didn't intend).
it was not one of those alibi questions "how's my product *extra-white-smile*? wanna buy cheap rdbms now?", but a serious attempt to keep in touch with people that really use and evaluate it. direct questions by technicians and not the usual marketing blah. i really was surprised.
oracle has some not too big focus on linux, when it's redhat or sles and will hardly ever satisfy the kind of linux lover running slackware, debian, gentoo or his own flavour.
they have never seen the benefits of linux in common because they never needed to.
for oracle, linux is unstable and inscalable if larry sais so until he sais anything different.
if larry hasn't said anything yet, don't really care about their statement in the meanwhile
in some countries, you get a drivers license (for anything you want) by simply setting the creation time of a file on a floppy disk. and who had not root access at school? it is cheating and it has to be punished, but i think this pupil just took the kirk solution for kobayashi maru test and it should not be too hard. the others that would have bought the test results are the real problem (and if they would not exist, it's questionable if he would have done it).
the thinkpad i got provided by some company doesn't even have windows keys, but two nice grey "switch terminal" buttons - just perfect for linux.
:)
it still was in the original packing and had no windows installed, only a small "designed for windows XP" sticker - that i kept for amusement.
and after all notebooks of other vendors (HP, compaq, acer) where i chose to run on linux, i'm really satisfied with my thinkpad.
of course, i had to tweak lots of things for myself, but very often i came across ibm e-mail addresses within the sources i had to compile. maybe they are not yet self confident enough to promote their linux support on their own (i mean, not any more their own but using their brand name) notebooks, but i'm already pleased by the few things they provide - it's at a good start at least.
tpbrc proposal:
THINKPAD firefox http://www.againsttcpa.com
as i started to love java, i was forced to learn PHP - and i did not like it, as any other script language, i considered it to be just another programmer newbee easy-to-learn and automate-anything-for-me language.
i argued against the decision of implementing a huge-scale web site in php, but i lost.
so i had to deal with it and read the php & zend engine sources (just a habit when dealing with an interpreter) and found out what it obviously is: easy to use glue code, where you have to know more about type casting, cost of algorithms and external functions than in java, if you want your huge project to perform. even the question where to hold what kind of resource (persistancy considerations), where to implement the abstractions gets a hard nut each time.
but after all, you get a very fast, perfectly scalable (just another box) application, no more dealing with fail-over stragegies - and if you don't have time to implement some special, hardly ever used feature, you get cheap external resources.
of all arguments i had against php, the strongest i still have is, that there are lots of developers that call themselves to know php and all you may expect them to code is an autoglobal using guestbook.
java is still great - and for some middle-scale applications, i still consider using it (besides all the other project dependent reasons for the language of choice) - but if it get's really big, i wouldn't use it any more since i experienced this kind of php usage.
(also proven by some j2ee to php migrations we did over the last 2 years)
you just talk about one of the big advantages of PHP - because it _does_ use this amount of external libraries. (and a pain in the ass for thee ones that have to update their special-purpose compilation right now)
/ext/gd of your php sources to see the crossing of bundled vs. external lib.
the wrappers are just needed for the zend engine to know how to map php-function parameters to the exports of any external lib (plus some extra things like constants or not too big abstractions for convenience and error handling).
your examples GD and MySQL happen to address two exceptions for good reasons. the history of GD deveolpment, bug-fixing and reachability of the maintainer just proposed this kind of bundle (e.g. the current gdlib has some build problems on solaris again)
take a look in the config.m4 in
for the bundled MySQL, there are some other reasons, leading from a tight cooperation of MySQL AB, Zend and some core developers.
but "php does not link enough external libs" is absolutely wrong.
the first announcement was on wednesday by stefan esser on bugtraq and full-disclosure. additional info came up later on, PHP 4.3.10 changelist and the according changes in CVS can be seen by everyone since then.
i do not think that any serious system administraton team gets to know this vulnerabilities via slashdot for the first time, the problem is, that compilation of all dependencies and deployment to all systems is a hard job to do from wednesday night to a reasonable end just before weekend.... in a week where lots of companies have their *drinks all inclusive* xmas events and the usual pre-vacation problems occur.
the w3schoole table is rather incomplete -
consider using php.net operators manual instead - there are more of them (but far away from maxing out ascii like perl 6 is about to do)
[i]"can i get one more of your sweets? i need 'em for this special move in street fighter"[/i] greets from the inventor of the "doom 2 walkthrough chewing gum" ;)
the ibm & ms divorce took place while there were 2 versions of os/2 in development v2 & v3 - and, primitively spoken (omitting some nasty war of roses), ms got parts of the v3 code base. to get a real os from it, they needed the cited programmers - and got vms developers.
this could have been a reason for windows NT starting with version 3.0 (and it's as likely as any other theory) and the mixture between ibm & vms way of problem-solution. a microsoft "own way" did not exist that time.
in the time we are talking about, microsoft was not that big knowledge center of development - but an interesting franchise strategy in the beginning of getting megalomaniac.
my connection is fast enough - but my gambing windows is always out of date (i don't use it more often than once a week on lan games) - so i download updates via linux and install from a fat32 partition before connecting a possibly risky windows to any network.
;)
as it's a notebook and a reinstalllation may be anywhere i have the time to play windows (train, plane, boring multihour kickoff meetings), it's better than a cd i'm sure to forget anywhere.
have you ever tried windows update under IE/wine? it's useless, but works
i think this argue was exaggerated. i'm used to synchronised movies & series and heard more than 5 different sean connery voices while edward norton is the same speaker as gauron in lord of the rings. it's not the sound of simpsons that makes me a fan , i'd really be disappointed if they ran out of yellow color ;-)
i'm used to the old installer and i learned to stop worrying and love ncurses frontends. despite of some naming conventions, i can't see much difference between freebsd, solaris and old debian installers.
ad knoppix:
once, i tried to do a knoppix auto-install (make own knoppix cd that installs itself on boot, starts networking (dhcp) and runs sshd for easier server installation (i forgot the special supersonic ibm kvm cable and was too curious about the new toy...) - it was unlikely to work, but in the end, there only was one single annoying point: knoppix didn't install the initrd image it uses itself from cd.
if a system runs from a knoppix cd, this does not mean, that it still runs after a knoppix-install, espacially with scsi inside...
driftnet similar, runs on linux and depends on - surprise - libpcap.
"networks became a lot more funny when me and my nic found the pleasures of promiscuousness"
each EU member has to merge the new restrictions (less cadmium, less lead, forced recycling,...) into their national law until 2005. it really starts on 1.1.2006 and concerns reduction of lead, mercury, cadmium & other heavy metal.
in summer 2001, philips semiconductors, infineon & some other committed to prduce nearly unleaded chips.
more info about EU RoHS (english version)
not never - but hey: i'm a debian user....
php, apache & tools i link against oracle - things i always compile myself in order to avoid problems. if you work in a heterogenous environment, this is one of the possibilities of keeping things even a lil' bit similar - and as a bonus, it's more stable. very common things like the most common gnu tools, i get from sunfreeware, debian, as ports or via cygwin - and i'm used to a lack of goodies (ahh- tar knows "z" here?). what really saves me a lot of time is a collection of shell scripts containing the configure strings and some written dependency trees (compiling the next missing lib on next term gets boring after a while - and too much coffee a day keeps the girlfriend away)
the chinese restaurant strikes back!