Ah well, looks like it's time to slap Ubuntu 6.06 LTS onto those old Fedora Core servers...
why?
theres plenty of good old redhat (6.2!) and fedora servers quietly humming away in the background doing their jobs quite nicely where i work ( and previous gigs too ).
sure, these are mainly internal systems that are used for in house stuf, but beyond the niceties of packaged installs of apps/utils, a adistro doesnt die just because 'official' support for them drops off over time.
my policy for fedora servers is even-numbers for public facing stuff ( ie: fc2, 4, 6), and for in house systems i'm mullung over skipping every second one of them, maybe more, we'll see over time.
even then, i'm more likely to look at the services they run ( and any security issues popping up around those specific versions ) than ditching a whole distribution. i'll typically build a system for particular purposes: mail or apache or java apps, so lifting off the service(s) and dropping them onto a new host with a shiny new fedora every 12-18 months is nothin'.
besides, being able to assess, rebuild and redeploy is half my job, so fedora keeps me in the loop nicely:)
i believe theres also third party services ( like progeny(sp?) ) which can provide paid errata support packages for aging systems.
after all, beyond support what are you paying oracle for that you cant get from postgres? and if oracle really is what you need, then the cost of oracle is gonna be waaaay more than an rhel license.
there ya go, your very first jsp page, set up in minutes.
first time i needed to get this going ( talking tomcat 3.x here, 6+ years ago? ) took less than an hour to google and implement. the bits and pieces have improved over time, and as the above should attest, it isnt rocket scientry these days.
setting up for a production system is only slightly more involved, and typically has more to do with the application itself than the container ( tomcat ), or the java runtime. most off-the-shelf java webapps ( see jira, confluence, wildfire) will have the step-by-step for the particulars of setting up the environment for the specific needs of the application deployment. say.. dont a lot of php apps also require tweaks to the environment? apache and or php.ini? and how many shared hosts allow you to set safe_mode?
developing a full blown cms? well, if the above is beyond you, then you should probably give up on php as well.
the real issue here is that there are very few hosting services who are in the market for java hosting. perhaps the demand isnt there? perhaps most shops deploying java webapps run it all in-house or manage their own hosted environments? java has been my bread and butter for years now, and never once have the customer ( or the company i've worked for) deployed to a managed host. always in house, providing both internal and external access.
one of the things about a php hosted environment is that the provider can lock your app down so it only runs as a low-privilleged user in a shared space ( http server with virtual hosts ). they can build a reasonable server for a grand or so, and charge N clients for shared access to it for bugger all. note here the key is 'shared environment'. the 'dedicated' packages that give you root shell access you are free to install and run whatever you like, in which case, see the above quick steps to setting up tomcat/apache.
applications written in java are typically more complex than php, but then again, they are more likely to perform more complex operations ( see banking & finance industries... and the lack of php therein ).
that said, you can make a java app as complex as you like, with all sorts of crazy requirements of the container, but that doesnt mean you have to. ( same goes for php y'know! )
i work in a tech company, mostly client services sorta stuff.
us tech types tend to get dropped into all sorts of different roles on client projects, engineer, developer, *ahem* architect, whatever.
so we dont put titles on our cards, means less to explain.
about 6 months ago, in another round of marketspeak genius, the 'business' types managed to snowball an email ( with html email sigs, 10k each) up to a 4.5 MB monster bickering over the difference between 'manager, client projects', and 'client projects manager'.
when really, all i need em to have is 'coffee bi-atch'
i've run linux on my desktop for years, and while i think its pretty easy to do everything i want to do, i concede its sometimes a bit much for joe user. not that any other OS is any different.
my ipod works fine when hooked up to my linux pc, although i thought i'd trashed it on the first connection ( hadnt realised it was hfs formatted, copied one track and yanked out the usb cable.. bad move)
what really surprises me though is that there arent more corporates or educational institutions rolling out linux on the desktop. i push it wherever i go.. not too hard, just open up my dell 9300 with fedora 5, and let em ask questions (ooooh... is that a mac? seems to come up a bit!)
surely something like ' hard to plug ipods into' would be a selling point?
but xen lets me set up/clone and configure clients that much faster.
guess it comes down to how and what you are developing, and for us, xen works very well. having vanilla installs for the testing cycle is probably a more accurate description of how we use the clients, rather than for direct coding.
and yeah, no acpi keeps xen off my laptop, but a no-X xenU domain is only ssh away:)
Yeah, for anyone who finds EO/WO, you think you've just hit paydirt as far as hooking up complex db's to a web front end.
i worked with it for a couple of years, but was a little constrained by the lack of Collections support, cos apple were really letting their jvm implementation fall behind ( ie: was still 1.3.x spec when i last did any real work in WO).
so finding cayenne with its use of standard java collections, as opposed to the NSArray, etc, was like winning the jackpot all over again. i did read some newer releases of WO had finally come round to the Collections thing, alas too late for me.
i've used cayenne in several projects over the last 12-18 months, and even the beta releases have been really stable. theres lots of good friendly input from the developers & users lists, and being able to point the modelling tool at an existing db and have a basic working model in minutes is pure gold, much like the first day working with the EOModeller:).
( ok, i'm anal when i build the db's, so i put em togher right to begin with... it doesnt always cover all the crazy db-as-a-big-bucket-o-data schemas i've seen..)
One thing i think they have definately improved on from EO is that the generated classes have a base class that carries all the DataObject methods, and a subclass that you implement your business logic in, so when you update the model and regenerate the class, you dont blow away your changes. cant remember how many times in my early days of EO that i overwrote all my custom code by regenerating the classes...argh.
so far i've used it with struts, spring and some straight up java back ends, and i hear folks who use the apache Tapestry project basically have an open source WebObjects stack, but i'm yet to give tapestry a whirl.
theres also a new 'client' tier feature that has some 'remote' object persistence feature, but havent had an excuse to delve into it.
so if you have a spare couple of hours, its all you'll need to read a little, click a little and be convinced...
( and no, i'm not involved directly with the project beyond submitting the odd bug/fix and comment on the list, just really impressed with what they've managed to produce...)
forget hibernate, i had the serious displeasure recently of listening to the self aggrandiosment of gavin king, and the way he prattles on you'd be forgiven for thinking the sun really did shine out his ass, and that the only way to do it was his way.
try cayenne, especially if you have a little WebObjects background. ( and even more so if you think hibernate is it and a bit )
anyway, back OT, +1 on the lashings with wet bamboo canes above please.
theres also a M$ provided way to do this, but i presume that means buying all the extras, etc.
unattended requires control over local dns/dhcp/samba/tftp on a preferably linux server box, and is sposed to work just as well with the windows server equivilents. ( and i presume the M$ one would have the same requirements..)
took me about 4 hrs from knowing nothing about it to having windows installed on a laptop with no OS, floppy or cd rom.
setting up different windowses is just a matter of copying the contents of the installation cd to a given directory, and i presume its probably the same for linux distros too ( although you'd probably attack it a different way for linux.)
also lets ya set up packages, and even modify the scripts to install different sets of packages as needed.
sure this was just to install to one machine, but now the setup is there, it can be used again, and again and again. ( used it to set up a vmware image right afterwards too, just for kicks..)
all thats required is a PXE compatible NIC, and that basically includes everything from about 2001 on.
still not sure i see the need to constantly 'rebuild' machines though. i shake my head when i see the windows sysadmins round the office constantly format, reinstalling, etc.
the argument is 'how can you afford to keep buying MS?'
i run a dozen or so linux 'production' servers. occasional housekeeping, sure, but apart from that, the maintenance is next to zilch.
we also have a couple of xen-linux development boxen, housing 20-30 virtual machines each ( not all of them running at the same ). this on el-cheapo desktop hardware. loaded to the tits with ram, but still, a couple hundred bucks a pop. cheap, quick, easy, reliable.
oh, yeah, this is in australia.
now the company has been merged. we have the other companies' network admin mandating 'new domain controller', 'sql server license', etc... ( not to mention the fucker wants to run the show from half way across the country over a shitty vpn link, roaming profiles and all)
the cfo has the decision in front of him: go the linux route ( 1 admin, several capable developers, no licensing/maintenance issues), or the MS route ( 3 admins, several capable developers, big licensing costs, one machine per purpose...)
which path do ya think he'll take?
back OT, NZ is a MS shop. they got in at the 2 main universities early, and wont let go. means all the kids learn visual whatever, makes em stick to windows. elsewhere, its java in the uni's, allowing coice of platforms.
as for the numbers.. man, if 18 % have adopted, and 11% are considering, and this represents 1/3 the numbers elsewher, then blow me dry and call me rusty
learn why management make the policy choices they do.
then have committee meetings to drag it out a little longer:)
actually, i found the double to be quite complimentary. the politics was geared toward policy processes, and while it may be pretty straight forward on the surface, it does change your perception about orders of heirarchies, etc.
Slashdot Mirror
Ah well, looks like it's time to slap Ubuntu 6.06 LTS onto those old Fedora Core servers...
:)
why?
theres plenty of good old redhat (6.2!) and fedora servers quietly humming away in the background doing their jobs quite nicely where i work ( and previous gigs too ).
sure, these are mainly internal systems that are used for in house stuf, but beyond the niceties of packaged installs of apps/utils, a adistro doesnt die just because 'official' support for them drops off over time.
my policy for fedora servers is even-numbers for public facing stuff ( ie: fc2, 4, 6), and for in house systems i'm mullung over skipping every second one of them, maybe more, we'll see over time.
even then, i'm more likely to look at the services they run ( and any security issues popping up around those specific versions ) than ditching a whole distribution. i'll typically build a system for particular purposes: mail or apache or java apps, so lifting off the service(s) and dropping them onto a new host with a shiny new fedora every 12-18 months is nothin'.
besides, being able to assess, rebuild and redeploy is half my job, so fedora keeps me in the loop nicely
i believe theres also third party services ( like progeny(sp?) ) which can provide paid errata support packages for aging systems.
how would oracle know?
:)
try log a support call
after all, beyond support what are you paying oracle for that you cant get from postgres? and if oracle really is what you need, then the cost of oracle is gonna be waaaay more than an rhel license.
web hosts may provide poor support, but claiming its hard to get tomcat and apache to work together is BS.
setting up tomcat and apache is dead simple.
$> sudo yum install tomcat5 tomcat5-webapps tomcat5-admin-webapps httpd
add proxy mapping for httpd -> tomcat on the ajp protocol in httpd.conf:
[ ProxyPass / ajp://localhost:8009 with httpd 2.2,
mod_jk for earlier httpd ]
add a html file to $TOMCAT_HOME/webapps/ROOT/, name it blah.jsp
browse to your http://webserver/blah.jsp
there ya go, your very first jsp page, set up in minutes.
first time i needed to get this going ( talking tomcat 3.x here, 6+ years ago? ) took less than an hour to google and implement. the bits and pieces have improved over time, and as the above should attest, it isnt rocket scientry these days.
setting up for a production system is only slightly more involved, and typically has more to do with the application itself than the container ( tomcat ), or the java runtime. most off-the-shelf java webapps ( see jira, confluence, wildfire) will have the step-by-step for the particulars of setting up the environment for the specific needs of the application deployment. say.. dont a lot of php apps also require tweaks to the environment? apache and or php.ini? and how many shared hosts allow you to set safe_mode?
developing a full blown cms? well, if the above is beyond you, then you should probably give up on php as well.
the real issue here is that there are very few hosting services who are in the market for java hosting. perhaps the demand isnt there? perhaps most shops deploying java webapps run it all in-house or manage their own hosted environments? java has been my bread and butter for years now, and never once have the customer ( or the company i've worked for) deployed to a managed host. always in house, providing both internal and external access.
one of the things about a php hosted environment is that the provider can lock your app down so it only runs as a low-privilleged user in a shared space ( http server with virtual hosts ). they can build a reasonable server for a grand or so, and charge N clients for shared access to it for bugger all. note here the key is 'shared environment'. the 'dedicated' packages that give you root shell access you are free to install and run whatever you like, in which case, see the above quick steps to setting up tomcat/apache.
applications written in java are typically more complex than php, but then again, they are more likely to perform more complex operations ( see banking & finance industries... and the lack of php therein ).
that said, you can make a java app as complex as you like, with all sorts of crazy requirements of the container, but that doesnt mean you have to. ( same goes for php y'know! )
http://www.google.com/linux
i'll agree to read this on condition the author agrees to stop beating his wife.
crappy languages like java?
as opposed to real-world languages like VB?
get me a long macchiatto, one sugar. :)
i work in a tech company, mostly client services sorta stuff.
us tech types tend to get dropped into all sorts of different roles on client projects, engineer, developer, *ahem* architect, whatever.
so we dont put titles on our cards, means less to explain.
about 6 months ago, in another round of marketspeak genius, the 'business' types managed to snowball an email ( with html email sigs, 10k each) up to a 4.5 MB monster bickering over the difference between 'manager, client projects', and 'client projects manager'.
when really, all i need em to have is 'coffee bi-atch'
i've run linux on my desktop for years, and while i think its pretty easy to do everything i want to do, i concede its sometimes a bit much for joe user. not that any other OS is any different.
my ipod works fine when hooked up to my linux pc, although i thought i'd trashed it on the first connection ( hadnt realised it was hfs formatted, copied one track and yanked out the usb cable.. bad move)
what really surprises me though is that there arent more corporates or educational institutions rolling out linux on the desktop. i push it wherever i go.. not too hard, just open up my dell 9300 with fedora 5, and let em ask questions (ooooh... is that a mac? seems to come up a bit!)
surely something like ' hard to plug ipods into' would be a selling point?
and just to flame myself:
:)
even better patches here:
http://www.openbsd.org/
better patches here:
http://fedora.redhat.com/
yes, vmware server rocks too.
:)
but xen lets me set up/clone and configure clients that much faster.
guess it comes down to how and what you are developing, and for us, xen works very well. having vanilla installs for the testing cycle is probably a more accurate description of how we use the clients, rather than for direct coding.
and yeah, no acpi keeps xen off my laptop, but a no-X xenU domain is only ssh away
i run about 40-50 xen clients on a handful of moderate server hosts.
perfect for dev work. i mean PERFECT
quickly reproducible, adjustable resourcing, and lets me give devs root acces on their own clients.
i presume the redhat dude meant was 'redhat isnt ready to commercially support xen'
----
what, read the article? pfft.
should be pretty easy in the current climate:
.. added bonus for the contractor is that word also removes accountability from public scrutiny.
just add the word 'terrorist' and an extra zero.
was there a reaganomics equivalent?
Yeah, for anyone who finds EO/WO, you think you've just hit paydirt as far as hooking up complex db's to a web front end.
:).
i worked with it for a couple of years, but was a little constrained by the lack of Collections support, cos apple were really letting their jvm implementation fall behind ( ie: was still 1.3.x spec when i last did any real work in WO).
so finding cayenne with its use of standard java collections, as opposed to the NSArray, etc, was like winning the jackpot all over again. i did read some newer releases of WO had finally come round to the Collections thing, alas too late for me.
i've used cayenne in several projects over the last 12-18 months, and even the beta releases have been really stable. theres lots of good friendly input from the developers & users lists, and being able to point the modelling tool at an existing db and have a basic working model in minutes is pure gold, much like the first day working with the EOModeller
( ok, i'm anal when i build the db's, so i put em togher right to begin with... it doesnt always cover all the crazy db-as-a-big-bucket-o-data schemas i've seen..)
One thing i think they have definately improved on from EO is that the generated classes have a base class that carries all the DataObject methods, and a subclass that you implement your business logic in, so when you update the model and regenerate the class, you dont blow away your changes. cant remember how many times in my early days of EO that i overwrote all my custom code by regenerating the classes...argh.
so far i've used it with struts, spring and some straight up java back ends, and i hear folks who use the apache Tapestry project basically have an open source WebObjects stack, but i'm yet to give tapestry a whirl.
theres also a new 'client' tier feature that has some 'remote' object persistence feature, but havent had an excuse to delve into it.
so if you have a spare couple of hours, its all you'll need to read a little, click a little and be convinced...
( and no, i'm not involved directly with the project beyond submitting the odd bug/fix and comment on the list, just really impressed with what they've managed to produce...)
cheers,
j
As for patentability? pfft.
yep,
.. shame apple dragged their feet on it.
EO/WO was the stuff back in the late 90's.
fortunately, these folks have taken the design principles and brought it all hurtling into modern java:
http://objectstyle.org/cayenne
its recently become an apache incubator project:
http://incubator.apache.org/projects/cayenne
forget hibernate, i had the serious displeasure recently of listening to the self aggrandiosment of gavin king, and the way he prattles on you'd be forgiven for thinking the sun really did shine out his ass, and that the only way to do it was his way.
try cayenne, especially if you have a little WebObjects background. ( and even more so if you think hibernate is it and a bit )
anyway, back OT, +1 on the lashings with wet bamboo canes above please.
http://unattended.sourceforge.net/
theres also a M$ provided way to do this, but i presume that means buying all the extras, etc.
unattended requires control over local dns/dhcp/samba/tftp on a preferably linux server box, and is sposed to work just as well with the windows server equivilents. ( and i presume the M$ one would have the same requirements..)
took me about 4 hrs from knowing nothing about it to having windows installed on a laptop with no OS, floppy or cd rom.
setting up different windowses is just a matter of copying the contents of the installation cd to a given directory, and i presume its probably the same for linux distros too ( although you'd probably attack it a different way for linux.)
also lets ya set up packages, and even modify the scripts to install different sets of packages as needed.
sure this was just to install to one machine, but now the setup is there, it can be used again, and again and again. ( used it to set up a vmware image right afterwards too, just for kicks..)
all thats required is a PXE compatible NIC, and that basically includes everything from about 2001 on.
still not sure i see the need to constantly 'rebuild' machines though. i shake my head when i see the windows sysadmins round the office constantly format, reinstalling, etc.
mate,
the argument is 'how can you afford to keep buying MS?'
i run a dozen or so linux 'production' servers. occasional housekeeping, sure, but apart from that, the maintenance is next to zilch.
we also have a couple of xen-linux development boxen, housing 20-30 virtual machines each ( not all of them running at the same ). this on el-cheapo desktop hardware. loaded to the tits with ram, but still, a couple hundred bucks a pop. cheap, quick, easy, reliable.
oh, yeah, this is in australia.
now the company has been merged. we have the other companies' network admin mandating 'new domain controller', 'sql server license', etc... ( not to mention the fucker wants to run the show from half way across the country over a shitty vpn link, roaming profiles and all)
the cfo has the decision in front of him: go the linux route ( 1 admin, several capable developers, no licensing/maintenance issues), or the MS route ( 3 admins, several capable developers, big licensing costs, one machine per purpose...)
which path do ya think he'll take?
back OT, NZ is a MS shop. they got in at the 2 main universities early, and wont let go. means all the kids learn visual whatever, makes em stick to windows. elsewhere, its java in the uni's, allowing coice of platforms.
as for the numbers.. man, if 18 % have adopted, and 11% are considering, and this represents 1/3 the numbers elsewher, then blow me dry and call me rusty
dude is just making up the numbers.
hehe,
:)
i think feynmann is also handing out the http headers for the new slash code:
X-BENDER
X-FRY
the new http headers :)
X-FRY
X-BENDER
hehe
.. and they have nothing to gain from happy workers skipping a couple of hours, that they'll probably make up again when they get in late?
:)
why doesnt this get factored in?
politics.
:)
learn why management make the policy choices they do.
then have committee meetings to drag it out a little longer
actually, i found the double to be quite complimentary. the politics was geared toward policy processes, and while it may be pretty straight forward on the surface, it does change your perception about orders of heirarchies, etc.
...in both US and Australia.
.. i thought australia IS part of the US now..
American wife
.. Now he warns me....
offer to quit.
:)
then sell back you services