Re: Tomcat 4.1-dev Host unpackWARs not working

[Glenn Nielsen <gl...@voyager.apg.more.net>]

Remy,

I posted that _OLD_ proposal from 3/13/01 as an example of how I 
documented the behaviour of tomcat for managing web apps.  It was
not meant as a current proposal.  Just as a place to start a discussion.

Could either you or Craig fully document what the expected behaviour is,
then perhaps we can have a good discussion on how it can be improved.

There are certain ways the current implementation works that I really
hate.  And has been a source of confusion for my customers.

To make my customer support easier Tomcat needs to have well defined,
well documented, and consistent behaviour when managing web apps.

A few more comments intermixed below.

Regards,

Glenn

Remy Maucherat wrote:
> 
> > [PROPOSAL]
> 
> -1 for this proposal (sorry, but I really *hate* it).
> 

Thanks, considering that a great deal of it was implemented by
me after I made it, and is still in place.

> To paraphrase a bit:
> If unpack="false" then unpack.
> If unpack="true" then unpack somewhere else.
> Otehrwise, both do the same.
> 
> The current behavior will hopefully encourage people not to rely on the
> filesystem and use more portable APIs (a good thing IMO).
> 
> -1 for removing expanded webapps on shutdown. You don't know if the user
> modified something, and would like to see its modifications survive the
> shutdown.

The old proposal didn't say that.

> At least unpack="false" currently makes things very clear, since the user
> can't modify anything without unpacking himself.
> 

Right, except that isn't the way my customers will use it.  They want
to be able to update files within the deployed app if needed to change
a JSP, etc.

> -1 for update as a new manager command. Its behavior seems highly
> unpredictable to me. remove + install seems better, so that the user knows
> what he's doing.
> 

I think update is an important feature to add.  A customer may need to update
the application but retain any data related to that application.  In that
case they wouldn't want to do an install/remove.  I think it can be done in
a predictable manner.  

> Note: remove/start/stop already exist.
> 

Right, because I added them after I made that proposal back on 3/13/01.

----------------------------------------------------------------------
Glenn Nielsen             glenn@more.net | /* Spelin donut madder    |
MOREnet System Programming               |  * if iz ina coment.      |
Missouri Research and Education Network  |  */                       |
----------------------------------------------------------------------

--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: Tomcat 4.1-dev Host unpackWARs not working

[Glenn Nielsen <gl...@voyager.apg.more.net>]

Remy Maucherat wrote:
> 
> > Remy,
> >
> > I posted that _OLD_ proposal from 3/13/01 as an example of how I
> > documented the behaviour of tomcat for managing web apps.  It was
> > not meant as a current proposal.  Just as a place to start a discussion.
> 
> Well, I didn't like it so far.
> 
> > Could either you or Craig fully document what the expected behaviour is,
> > then perhaps we can have a good discussion on how it can be improved.
> 
> The thing is that I don't think it can be improved without introducing some
> unpredictability.
> 
> > > To paraphrase a bit:
> > > If unpack="false" then unpack.
> > > If unpack="true" then unpack somewhere else.
> > > Otehrwise, both do the same.
> > >
> > > The current behavior will hopefully encourage people not to rely on the
> > > filesystem and use more portable APIs (a good thing IMO).
> > >
> > > -1 for removing expanded webapps on shutdown. You don't know if the user
> > > modified something, and would like to see its modifications survive the
> > > shutdown.
> >
> > The old proposal didn't say that.
> >
> > > At least unpack="false" currently makes things very clear, since the
> user
> > > can't modify anything without unpacking himself.
> > >
> >
> > Right, except that isn't the way my customers will use it.  They want
> > to be able to update files within the deployed app if needed to change
> > a JSP, etc.
> 
> I won't let them if unpack="false" (the WAR will be packed). Otherwise, set
> it to true (the WAR will be unpacked). It seems a bit obvious ...
> 

Except that isn't true all the time. In some cases the manager will override
the Host configuration setting of unpackWARs="true" and _not_ unpack them.

> > > -1 for update as a new manager command. Its behavior seems highly
> > > unpredictable to me. remove + install seems better, so that the user
> knows
> > > what he's doing.
> > >
> >
> > I think update is an important feature to add.  A customer may need to
> update
> > the application but retain any data related to that application.  In that
> > case they wouldn't want to do an install/remove.  I think it can be done
> in
> > a predictable manner.
> 
> I doubt it at this point. I'm not in favor of adding any intelligence and
> do-stuff-behind-your-back features in the deployer.
> 

Its not do-stuff-behind-your-back if how it does an update is well documented.

Glenn

----------------------------------------------------------------------
Glenn Nielsen             glenn@more.net | /* Spelin donut madder    |
MOREnet System Programming               |  * if iz ina coment.      |
Missouri Research and Education Network  |  */                       |
----------------------------------------------------------------------

--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: Tomcat 4.1-dev Host unpackWARs not working

[Remy Maucherat <re...@apache.org>]

> Remy,
>
> I posted that _OLD_ proposal from 3/13/01 as an example of how I
> documented the behaviour of tomcat for managing web apps.  It was
> not meant as a current proposal.  Just as a place to start a discussion.

Well, I didn't like it so far.

> Could either you or Craig fully document what the expected behaviour is,
> then perhaps we can have a good discussion on how it can be improved.

The thing is that I don't think it can be improved without introducing some
unpredictability.

> > To paraphrase a bit:
> > If unpack="false" then unpack.
> > If unpack="true" then unpack somewhere else.
> > Otehrwise, both do the same.
> >
> > The current behavior will hopefully encourage people not to rely on the
> > filesystem and use more portable APIs (a good thing IMO).
> >
> > -1 for removing expanded webapps on shutdown. You don't know if the user
> > modified something, and would like to see its modifications survive the
> > shutdown.
>
> The old proposal didn't say that.
>
> > At least unpack="false" currently makes things very clear, since the
user
> > can't modify anything without unpacking himself.
> >
>
> Right, except that isn't the way my customers will use it.  They want
> to be able to update files within the deployed app if needed to change
> a JSP, etc.

I won't let them if unpack="false" (the WAR will be packed). Otherwise, set
it to true (the WAR will be unpacked). It seems a bit obvious ...

> > -1 for update as a new manager command. Its behavior seems highly
> > unpredictable to me. remove + install seems better, so that the user
knows
> > what he's doing.
> >
>
> I think update is an important feature to add.  A customer may need to
update
> the application but retain any data related to that application.  In that
> case they wouldn't want to do an install/remove.  I think it can be done
in
> a predictable manner.

I doubt it at this point. I'm not in favor of adding any intelligence and
do-stuff-behind-your-back features in the deployer.

Remy


--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>