Scientific publishing is a racket similar to the mafia. The only difference is that scientific publishers don't kill you with a bullet; it's just if you don't cooperate and put in hundreds of thousands of dollars worth of specialized, unpaid labor, your scientific career is over.
I agree, it is a bogus process. You work to get your paper out, get half of them rejected by idiot reviewers (who are unpaid by the publisher), then when they are accepted the publisher takes the copyright and makes everyone pay to see them.
The web is a great threat to this whole business model, though personal-web-site publishing lacks rigorous peer review. We need to use MIT d-space (slashdot passim)and the like to put all academic work on line in a rigorous manner, and we can take the publishers out the loop.
If you ever write a book (it's worth trying,, just not while having a full time job as a software developer), you'll understand why books are written as if the author was asleep; I think I only got 6 hours sleep for 4 months in row, it was brutal.
I can also see your prejudice -I've read the other two ant books and I dont thing they bring enough to the table above the Ant docs. They have a better introductory guide, but once you know ant, that's no longer useful. We wrote ours assuming you have the ant docs open in a different pane of mozilla from the JDK docs, and added stuff that isnt in there. And we fixed ant as we went along when we found problems: you can guess when we wrote chapters by looking at the ant bugzilla entries.
My recommendation would be have a look at your local bookshop, sit down in a corner with a coffee and your laptop, and read through bits of it. Pick a chapter related to a project you are working on. Up on the manning site we have posted a couple of chapters. Ch4 is about junit; I assume you know that already. Chapter15 shows how to build apache axis based web services and how to interop test against.net. In the bookshop you should also scope out ch11, Xdoclet, ch12, ch13 'xml', and maybe ch18, production side deployment. Its not often a book on java has a chapter criticising the many j2ee implementations for all being different, or operations teams who cal up engineers at 3am whenever they get an error message they dont understand, but in that chapter we do it:)
-steve
(pet peeve: books about OSS projects that dont add to the docs and whose authors didnt contribute anything to the project, not even bugreps)
DTD: create one with , but it has problems because ant is so dynamic. There are discussions about having an XML descriptor for each task so that we could really get it right. You still have to deal with the problem that you could use a task in the same file you compile it.
An ant debugger. Maybe, as part of better GUI tools. Maybe we should add a debug interface to the engine and let people (you?) code one. I usually use tests and checks to verify that jars I'm using exist, to list properties, and the infamous -verbose and -debug switches.
Your rollback need shows a limitation of ant: it isnt an installer, not a real one. We use it a lot for server side deployment; in that context my buil file backs off the old war/ear for emergency restoration later. Because the big problems I have is not that the deploy code doesnt work, but what I deployed is broken...
-Steve Loughran, co-author of Java Development with Ant.
It is meant to be a declarative language to be interpreted by an engine; there have been multiple implementations of the ant engine (there are two in ant CVS), which can take the same XML declaration and build things because of it.
XML is a different issue. Yes it can be ugly, but compared to the secret tab characters in make, it is better. You can use XML editors (I use jedit) to edit it, and its easy to generate ant build files from higher level specification files using XSLT and other XML processing chains. Or vice versa: there are tools to take ant files and parse and navigate them, such as the IDEA, forte and jedit and plugins, or the vizant ant project visualiser.
Also look at maven which is a high level wrapper around ant for real project work.
But thank you for your AAP pointer; i will look at it.
wow, mod this person up. nobody has said nice things about me on./ before.
Of my presentations, my current favourite is actually The Wondrous Curse of Interoperability , where I get to criticise everyone's SOAP implementation, even the one I work, on, Apache Axis. Plus it has excellent mountaineering photos.
You still might want to look at some of the advanced subjects, such as how to compile jni native code using , then run junit tests against it, how to set up cruise control or the gump to rebuild your code hourly, how to build C# clients to interop test your web service, etc.
If we'd stopped at the first section and added a rewritten task reference you'd be making a valid point. But we wrote the book as a supplement to the docs. After all, our hand is in the docs too, from Ant in anger or Ant Task guidelines (both mine), to all the little corrections we made to the ant docs when we wrote the book and disovered errors.
So use the docs -I think they are really good compared to many other OSS projects, and I am glad you like them. But I think you should still find value in Java Development with Ant. Erik and I really pushed the envelope in section two, doing stuff nobody had done in ant before.
This isnt an ant bug per-se. It is a java bug: no way of getting the permissions. "You have no use for that information", say Sun:-( . Actually, ant's zip task doesnt store permissions, so if it could detect them, then a bug would surface.
If you use the task, you can specify the perms on a , so the permissions on your tar distro is all set up. This even lets you create proper tar files on a windows box that doesnt understand permissions at all.
-Steve Loughran, co-author Java Development with Ant
Really you mean Make+Autoconf+Unix is great, and who can argue with that?
But ant is nice because it is fully x-platform, from MacOS X, OS/400, Netware, Linux, Solaris, NT. Even Win9x, though that raises too many support calls.
Also it is nice because makes dependency model is tied to the local filesystem: its notion of 'needs doing' means dest file is older than source file.
Ant offloads dependencies to tasks, and each task has its own logic. The command talks stright to the database, knows about timestamps of dependencies over the wire, etc.
Sometimes make is better, because it adds dependencies to those dumb unix tools that just 'do' things. Its crufty to do the same in ant with tests and conditional targets. The other failing of Ant is that it doesnt necessarily scale as well as gnu make+autoconf does. We will have to work on that,
-Steve Loughran, co-author Java Development with Ant
By the time you've read the first eight chapters you ready to either write your own build files, up to and including junit testing and deployment. So you can stop reading there if you want. But at that point you will probably need to look at one or two of the advanced topics, depending on what you want. Doing webapps? read ch12, Doing EJB? Ch 14. Soap? chapter 15.
The final section, extending ant sems a bit off topic to pure ant consumers, but we wanted to show it is easy to do, when Ant doesnt meet all your needs. Its easy to integrate other java code into Ant.
I Think there is still scope for a single 'Ant Internals' book that focuses on the code, explains the IntrospectionHelper, how tasks work, covers task testing, etc etc. But that is a niche product, you need to be a core ant developer to want that book. And we have the mail list and the source in that situation.
You are right; Ant does timestamp based comparison before deciding what to send to javac, or jikes.
But it does not (by default) do full dependency checking to force rebuild a file when a class it imports changes. The Task does that, (something we cover in chapter, 10 BTW). Even that task, which looks through the.class file to discover the imports, doesnt handle import of constants, as that isnt logged in the class files.
There is a lot to be said for clean builds using jikes, as that is reliable and fast.
-Steve Loughran, co-author Java Development with Ant
Maybe I should write to a record company telling them that their copy-protected CD has a scratch on, and could I have a replacement for the cost of a postage stamp? I bet I'd get a rude response too.
That's what comes of BEA and WebSphere pricings; if they'd used jboss or HP-app server (still a free download, I believe), app server cost becomes zero, leaving only hardware, OS and database. Move to hypersonic SQL or postgres to take oracle out the loop and you start to get economic performance. Sure, per-cpu perf may be less, but what if all those CPUs are just blade mounted PIII-8000 parts running your favourite linux distro?
NB, after development costs, the biggest expenses in my last projects were operational costs and hardware depreciation. This review didnt look at either of those, but everyone I know is scared of IIS management, so its operational cost is probably higher.
yes, comparing a lightweight struts/JDO impl running on Jetty would be more interesting.
I think sun have got into a mess pushing EJB as the be-all and end-all of server side java coding. I dont know whether that was because they wanted to enable client server apps using the EJB api too, justify to customers the purchase of premium j2ee servers over free servlet engines, or encourage sales of multiway solaris boxes.
But because EJB is wierd, ugly and so limited, I dont think it is the right design for many apps. The fact that it can take weeks of 'tuning' the app server to get acceptable performance out of it is a fatal flaw in its own right. nobody has time to do that.
I dunno about that. A fair few of the corner case ORA books are pretty mediocre. Some of them are great -its these that create the reputation that other volumes just use, rather than expand.
To be fair, the great:good:mediocre ratio for oreilly is better than most other publishers.
I agree that it is hard to automate UI Testing; you cant automate usability tests. But it is good for basic regression testing: making sure links are still there, etc. just think how much better the web would be if every site, every night, ran a regression test to verify that their local site worked and the external links werent 404-ing
are you sure about the lack of javascript support? I am running the latest version and every test it complains that rhino.jar isnt found, so scripting is off. I wonder if someone hasn't just added it...
-a good trick with httpunit is to run it under ant's then the results in a summary page -this never fails to impress.
http unit is good for probing web pages, parsing content, verifying links off it work. Cactus is more for testing the classes behind the web page, if that makes sense, a kind of RPC back door into the beans.
httpunit works against any web site, not just java, it just gets, posts, and analyses the results.
Best book on httpunit is 'java tools for extreme programming'; 'java development with ant' also looks at it from the context of automated build and test processes. you can see that book applied in apache axis, under test/httpunit.
well, Search for 'Sam Ruby' on google to get his intro, and d/l chapter 15 of my ant book (http://manning.com/antbook) for the practicalities of doing apache SOAP work. Best book on Axis I've seen is the Sams press one; the Wrox book on axis is a bit more up to date (I havent seen it).
To bring up axis you need to get axis.war into your app server, make sure happyaxis.jsp is happy. Then you can add new services by writing java files with.jws extension and dropping them into the expanded WAR. Each file becomes an endpoint where methods can be posted to. postRC1.0 (since last friday), axis has supported GET too.
well, ASP.net is *nominally* better than jsp; at least with the design UI that VS.net gives you; asp.net are roughly equivalent in functionality to taglibs.
But jsp pages are nasty for complex code, so are asp.net pages. And the execution of some of the ASP.net controls take up a lot of bandwidth; the calendar control is an example.
ASP.NET's web service model is good; it is very easy to export methods from a class as SOAP actions against a URL. But to build and deploy a production quality, high availablity web service is always a management and config nightmare, and there java still has the edge over IIS+WinNT.
they also want to get into the biz of selling music and movies too, and dont want the media companies to lock out their options.
But they must know what sells broadband to the masses today, and they must also fear what kind of support calls they will be getting if ever the RIAA do start attacking P2P networks. Nobody wants support calls, especially of the type 'All my files are missing, replaced with the text 'RIAAMPAA' again and again and again'.
This has been known about since 2000 at least; we used to have endless discussions over the fact that geckos have the impressive ability to stick to ceilings in a vacuum, discussions on topics such as:
a) how did they find out the details? Did it involve a research assistant, a glass container, a vacuum pump and a large supply of geckos?
b) How did Geckos evolve this feature? Are geckos secretly descended from a life form that can stick to the outside of space craft?
c) Alternatively, does this prove that creatures are designed rather than evolved, and the design process is a bit more like the PhD process than anything else; some little godling spends millenia working on geckos in order to submit some paper 'An alternative mechanism for achieving stickiness in creatures' only to have it discredited by a board of professors who have always used suction and thats how they believe all creatures should stick.
I can see political spam changing if things take off. At the very least, spammers could hide their stuff as politics if political spam were legal.
"I was a lowly senator, unable to service my hot young interns (see them now at hot-young-interns-in-the-senate.com), until I bought this herbal extract which works like Viagra for less. Buy it here...for every $10 spent, $1 goes to my releection campaign".
Or
"dear sirs, I am writing to you in utmost confidence, as a former republican congressman of Texas, now in exile in Sierra Leone. An aide of mine, on loan from Enron, has the information needed to get at $17 million of Enron investment information from an offshore account in Nigeria. I need your help to get this money, with which I will take a small percentage to pay for TV advertisements"...
Slashdot Mirror
ok, now I remember you vaguely. Glad you like the book, and it was a most excellent conference, was it not?
I agree, it is a bogus process. You work to get your paper out, get half of them rejected by idiot reviewers (who are unpaid by the publisher), then when they are accepted the publisher takes the copyright and makes everyone pay to see them.
The web is a great threat to this whole business model, though personal-web-site publishing lacks rigorous peer review. We need to use MIT d-space (slashdot passim)and the like to put all academic work on line in a rigorous manner, and we can take the publishers out the loop.
If you ever write a book (it's worth trying,, just not while having a full time job as a software developer), you'll understand why books are written as if the author was asleep; I think I only got 6 hours sleep for 4 months in row, it was brutal.
.net. In the bookshop you should also scope out ch11, Xdoclet, ch12, ch13 'xml', and maybe ch18, production side deployment. Its not often a book on java has a chapter criticising the many j2ee implementations for all being different, or operations teams who cal up engineers at 3am whenever they get an error message they dont understand, but in that chapter we do it :)
I can also see your prejudice -I've read the other two ant books and I dont thing they bring enough to the table above the Ant docs. They have a better introductory guide, but once you know ant, that's no longer useful. We wrote ours assuming you have the ant docs open in a different pane of mozilla from the JDK docs, and added stuff that isnt in there. And we fixed ant as we went along when we found problems: you can guess when we wrote chapters by looking at the ant bugzilla entries.
My recommendation would be have a look at your local bookshop, sit down in a corner with a coffee and your laptop, and read through bits of it. Pick a chapter related to a project you are working on.
Up on the manning site we have posted a couple of chapters. Ch4 is about junit; I assume you know that already. Chapter15 shows how to build apache axis based web services and how to interop test against
-steve
(pet peeve: books about OSS projects that dont add to the docs and whose authors didnt contribute anything to the project, not even bugreps)
DTD: create one with , but it has problems because ant is so dynamic. There are discussions about having an XML descriptor for each task so that we could really get it right. You still have to deal with the problem that you could use a task in the same file you compile it.
An ant debugger. Maybe, as part of better GUI tools. Maybe we should add a debug interface to the engine and let people (you?) code one. I usually use tests and checks to verify that jars I'm using exist, to list properties, and the infamous -verbose and -debug switches.
Your rollback need shows a limitation of ant: it isnt an installer, not a real one. We use it a lot for server side deployment; in that context my buil file backs off the old war/ear for emergency restoration later. Because the big problems I have is not that the deploy code doesnt work, but what I deployed is broken...
-Steve Loughran, co-author of Java Development with Ant.
Ant is not meant be a scripting language.
It is meant to be a declarative language to be interpreted by an engine; there have been multiple implementations of the ant engine (there are two in ant CVS), which can take the same XML declaration and build things because of it.
XML is a different issue. Yes it can be ugly, but compared to the secret tab characters in make, it is better. You can use XML editors (I use jedit) to edit it, and its easy to generate ant build files from higher level specification files using XSLT and other XML processing chains. Or vice versa: there are tools to take ant files and parse and navigate them, such as the IDEA, forte and jedit and plugins, or the vizant ant project visualiser.
Also look at maven
which is a high level wrapper around ant for real project work.
But thank you for your AAP pointer; i will look at it.
Steve Lougran.
wow, mod this person up. nobody has said nice things about me on ./ before.
Of my presentations, my current favourite is actually
The Wondrous Curse of Interoperability , where I get to criticise everyone's SOAP implementation, even the one I work, on, Apache Axis. Plus it has excellent mountaineering photos.
You can skip the first eight chapters then
You still might want to look at some of the advanced subjects, such as how to compile jni native code using , then run junit tests against it, how to set up cruise control or the gump to rebuild your code hourly, how to build C# clients to interop test your web service, etc.
If we'd stopped at the first section and added a rewritten task reference you'd be making a valid point. But we wrote the book as a supplement to the docs. After all, our hand is in the docs too, from Ant in anger or Ant Task guidelines (both mine), to all the little corrections we made to the ant docs when we wrote the book and disovered errors.
So use the docs -I think they are really good compared to many other OSS projects, and I am glad you like them. But I think you should still find value in Java Development with Ant. Erik and I really pushed the envelope in section two, doing stuff nobody had done in ant before.
This isnt an ant bug per-se. It is a java bug: no way of getting the permissions. "You have no use for that information", say Sun :-( . Actually, ant's zip task doesnt store permissions, so if it could detect them, then a bug would surface.
If you use the task, you can specify the perms on a , so the permissions on your tar distro is all set up. This even lets you create proper tar files on a windows box that doesnt understand permissions at all.
-Steve Loughran, co-author Java Development with Ant
Really you mean Make+Autoconf+Unix is great, and who can argue with that?
But ant is nice because it is fully x-platform, from MacOS X, OS/400, Netware, Linux, Solaris, NT. Even Win9x, though that raises too many support calls.
Also it is nice because makes dependency model is tied to the local filesystem: its notion of 'needs doing' means dest file is older than source file.
Ant offloads dependencies to tasks, and each task has its own logic. The command talks stright to the database, knows about timestamps of dependencies over the wire, etc.
Sometimes make is better, because it adds dependencies to those dumb unix tools that just 'do' things. Its crufty to do the same in ant with tests and conditional targets. The other failing of Ant is that it doesnt necessarily scale as well as gnu make+autoconf does. We will have to work on that,
-Steve Loughran, co-author Java Development with Ant
I must take exception to your claim.
By the time you've read the first eight chapters you ready to either write your own build files, up to and including junit testing and deployment. So you can stop reading there if you want. But at that point you will probably need to look at one or two of the advanced topics, depending on what you want. Doing webapps? read ch12, Doing EJB? Ch 14. Soap? chapter 15.
The final section, extending ant sems a bit off topic to pure ant consumers, but we wanted to show it is easy to do, when Ant doesnt meet all your needs. Its easy to integrate other java code into Ant.
I Think there is still scope for a single 'Ant Internals' book that focuses on the code, explains the IntrospectionHelper, how tasks work, covers task testing, etc etc. But that is a niche product, you need to be a core ant developer to want that book. And we have the mail list and the source in that situation.
You are right; Ant does timestamp based comparison before deciding what to send to javac, or jikes.
.class file to discover the imports, doesnt handle import of constants, as that isnt logged in the class files.
But it does not (by default) do full dependency checking to force rebuild a file when a class it imports changes. The Task does that, (something we cover in chapter, 10 BTW). Even that task, which looks through the
There is a lot to be said for clean builds using jikes, as that is reliable and fast.
-Steve Loughran, co-author Java Development with Ant
Yeah, I like my music backups too.
Maybe I should write to a record company telling them that their copy-protected CD has a scratch on, and could I have a replacement for the cost of a postage stamp? I bet I'd get a rude response too.
That's what comes of BEA and WebSphere pricings; if they'd used jboss or HP-app server (still a free download, I believe), app server cost becomes zero, leaving only hardware, OS and database. Move to hypersonic SQL or postgres to take oracle out the loop and you start to get economic performance. Sure, per-cpu perf may be less, but what if all those CPUs are just blade mounted PIII-8000 parts running your favourite linux distro?
NB, after development costs, the biggest expenses in my last projects were operational costs and hardware depreciation. This review didnt look at either of those, but everyone I know is scared of IIS management, so its operational cost is probably higher.
yes, comparing a lightweight struts/JDO impl running on Jetty would be more interesting.
I think sun have got into a mess pushing EJB as the be-all and end-all of server side java coding. I dont know whether that was because they wanted to enable client server apps using the EJB api too, justify to customers the purchase of premium j2ee servers over free servlet engines, or encourage sales of multiway solaris boxes.
But because EJB is wierd, ugly and so limited, I dont think it is the right design for many apps. The fact that it can take weeks of 'tuning' the app server to get acceptable performance out of it is a fatal flaw in its own right. nobody has time to do that.
>Hard to go wrong buying an O'Reilly book
I dunno about that. A fair few of the corner case ORA books are pretty mediocre. Some of them are great -its these that create the reputation that other volumes just use, rather than expand.
To be fair, the great:good:mediocre ratio for oreilly is better than most other publishers.
hey, have you tried an early access version of IDEA3.0? Great JSP editing, even refactoring inside jsp code -not that that is ever a good thing.
good point about data binding though. JSTL is not the answer, at least to me.
I agree that it is hard to automate UI Testing; you cant automate usability tests. But it is good for basic regression testing: making sure links are still there, etc. just think how much better the web would be if every site, every night, ran a regression test to verify that their local site worked and the external links werent 404-ing
are you sure about the lack of javascript support? I am running the latest version and every test it complains that rhino.jar isnt found, so scripting is off. I wonder if someone hasn't just added it...
-a good trick with httpunit is to run it under ant's then the results in a summary page -this never fails to impress.
http unit is good for probing web pages, parsing content, verifying links off it work. Cactus is more for testing the classes behind the web page, if that makes sense, a kind of RPC back door into the beans.
httpunit works against any web site, not just java, it just gets, posts, and analyses the results.
Best book on httpunit is 'java tools for extreme programming'; 'java development with ant' also looks at it from the context of automated build and test processes. you can see that book applied in apache axis, under test/httpunit.
-steve (who co-authored java dev with ant)
well, Search for 'Sam Ruby' on google to get his intro, and d/l chapter 15 of my ant book (http://manning.com/antbook) for the practicalities of doing apache SOAP work. Best book on Axis I've seen is the Sams press one; the Wrox book on axis is a bit more up to date (I havent seen it).
.jws extension and dropping them into the expanded WAR. Each file becomes an endpoint where methods can be posted to. postRC1.0 (since last friday), axis has supported GET too.
To bring up axis you need to get axis.war into your app server, make sure happyaxis.jsp is happy. Then you can add new services by writing java files with
well, ASP.net is *nominally* better than jsp; at least with the design UI that VS.net gives you; asp.net are roughly equivalent in functionality to taglibs.
But jsp pages are nasty for complex code, so are asp.net pages. And the execution of some of the ASP.net controls take up a lot of bandwidth; the calendar control is an example.
ASP.NET's web service model is good; it is very easy to export methods from a class as SOAP actions against a URL. But to build and deploy a production quality, high availablity web service is always a management and config nightmare, and there java still has the edge over IIS+WinNT.
yeah, maybe they will start advertising this fact...verizon man walking around with a laptop full of mp3s going "Can you see my files now?"
they also want to get into the biz of selling music and movies too, and dont want the media companies to lock out their options.
But they must know what sells broadband to the masses today, and they must also fear what kind of support calls they will be getting if ever the RIAA do start attacking P2P networks. Nobody wants support calls, especially of the type 'All my files are missing, replaced with the text 'RIAAMPAA' again and again and again'.
This has been known about since 2000 at least; we used to have endless discussions over the fact that geckos have the impressive ability to stick to ceilings in a vacuum, discussions on topics such as:
a) how did they find out the details? Did it involve a research assistant, a glass container, a vacuum pump and a large supply of geckos?
b) How did Geckos evolve this feature? Are geckos secretly descended from a life form that can stick to the outside of space craft?
c) Alternatively, does this prove that creatures are designed rather than evolved, and the design process is a bit more like the PhD process than anything else; some little godling spends millenia working on geckos in order to submit some paper 'An alternative mechanism for achieving stickiness in creatures' only to have it discredited by a board of professors who have always used suction and thats how they believe all creatures should stick.
I can see political spam changing if things take off. At the very least, spammers could hide their stuff as politics if political spam were legal.
"I was a lowly senator, unable to service my hot young interns (see them now at hot-young-interns-in-the-senate.com), until I bought this herbal extract which works like Viagra for less. Buy it here...for every $10 spent, $1 goes to my releection campaign".
Or
"dear sirs,
I am writing to you in utmost confidence, as a former republican congressman of Texas, now in exile in Sierra Leone. An aide of mine, on loan from Enron, has the information needed to get at $17 million of Enron investment information from an offshore account in Nigeria. I need your help to get this money, with which I will take a small percentage to pay for TV advertisements"
etc.