Home » Archived » B3 » API baseline - relevant for b3
|
Re: API baseline - relevant for b3 [message #488242 is a reply to message #488235] |
Sat, 26 September 2009 15:35 |
|
On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
> have any bearing on b3 ?
>
> (Ability to set API baseline).
>
Yes, I think it will be relevant.
An API baseline (AB for short) is very similar to a target platform. The difference is that a TP can be updated several
times during a development cycle but the AB is often set right at the beginning of that cycle and then used as a
verification mechanism to assert that nothing deviates from that AB. So the AB verifies both the workspace and the TP.
Presently, the AB is simply a workspace global named entry that appoints a folder with plug-in's in it. Very similar to
the TP but more limited in how it can be described. And like the current PDE TP, it's OSGi specific. I don't think it
has to be though.
A workspace can have many AB's but only one is active at any one time. The compiler uses both the AB and the TP while
compiling.
- thomas
|
|
|
Re: API baseline - relevant for b3 [message #488286 is a reply to message #488242] |
Sun, 27 September 2009 15:49 |
Henrik Lindberg Messages: 2509 Registered: July 2009 |
Senior Member |
|
|
So, does this require new materializer, or is it the p2 based
materializer that does this?
This leads to a different question - When materializing to a TP, or a
AB, I assume that the materialization itself does not make it the active
one (unless of course the materialization is performed to the location
of the active TP/AB).
I can imagine specifying symbolic locations for "active TP", and "active
AB" - but this seems a bit dangerous.
So, how can these settings be specified? In a query? What should happen
in UI mode? Maybe I don't want to switch at all (I am populating my
workspace using several queries, and I am going to update AB etc.).
Should it popup questions before switching? In headless mode it should
probably just switch if so directed?
I also think this leads to issues of materializing/provisioning of
preferecens/settings.
- henrik
Thomas Hallgren wrote:
>
> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>> have any bearing on b3 ?
>>
>> (Ability to set API baseline).
>>
> Yes, I think it will be relevant.
>
> An API baseline (AB for short) is very similar to a target platform. The
> difference is that a TP can be updated several times during a
> development cycle but the AB is often set right at the beginning of that
> cycle and then used as a verification mechanism to assert that nothing
> deviates from that AB. So the AB verifies both the workspace and the TP.
>
> Presently, the AB is simply a workspace global named entry that appoints
> a folder with plug-in's in it. Very similar to the TP but more limited
> in how it can be described. And like the current PDE TP, it's OSGi
> specific. I don't think it has to be though.
>
> A workspace can have many AB's but only one is active at any one time.
> The compiler uses both the AB and the TP while compiling.
>
> - thomas
|
|
|
Re: API baseline - relevant for b3 [message #488287 is a reply to message #488286] |
Sun, 27 September 2009 17:46 |
|
Hi Henrik,
You seem to be pointing to one of the weak areas of the API Tools
infrastructure: modularity of the API baseline (AB). I understand that
this is not primarily a concern for Buckminster or b3, but maybe in this
forum we can gather enough "weight" to convince the API Tools team that
AB modularity is not less important than TP modularity. I wonder if they
were willing to consider the reuse of the existing TP infrastructure
instead of keeping their own, unmodular solution.
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
Henrik Lindberg schrieb:
> So, does this require new materializer, or is it the p2 based
> materializer that does this?
>
> This leads to a different question - When materializing to a TP, or a
> AB, I assume that the materialization itself does not make it the
> active one (unless of course the materialization is performed to the
> location of the active TP/AB).
>
> I can imagine specifying symbolic locations for "active TP", and
> "active AB" - but this seems a bit dangerous.
>
> So, how can these settings be specified? In a query? What should
> happen in UI mode? Maybe I don't want to switch at all (I am
> populating my workspace using several queries, and I am going to
> update AB etc.). Should it popup questions before switching? In
> headless mode it should probably just switch if so directed?
>
> I also think this leads to issues of materializing/provisioning of
> preferecens/settings.
>
> - henrik
>
>
> Thomas Hallgren wrote:
>>
>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>> have any bearing on b3 ?
>>>
>>> (Ability to set API baseline).
>>>
>> Yes, I think it will be relevant.
>>
>> An API baseline (AB for short) is very similar to a target platform.
>> The difference is that a TP can be updated several times during a
>> development cycle but the AB is often set right at the beginning of
>> that cycle and then used as a verification mechanism to assert that
>> nothing deviates from that AB. So the AB verifies both the workspace
>> and the TP.
>>
>> Presently, the AB is simply a workspace global named entry that
>> appoints a folder with plug-in's in it. Very similar to the TP but
>> more limited in how it can be described. And like the current PDE TP,
>> it's OSGi specific. I don't think it has to be though.
>>
>> A workspace can have many AB's but only one is active at any one
>> time. The compiler uses both the AB and the TP while compiling.
>>
>> - thomas
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
Re: API baseline - relevant for b3 [message #488300 is a reply to message #488287] |
Sun, 27 September 2009 20:20 |
|
I agree with Ekie. I just don't understand why two different concepts exists that have that much overlap. I.e. Why isn't
an AB just another TP? Or to rephrase it, why not define an AB by appointing a TP?
- thomas
On 09/27/2009 07:46 PM, Eike Stepper wrote:
> Hi Henrik,
>
> You seem to be pointing to one of the weak areas of the API Tools
> infrastructure: modularity of the API baseline (AB). I understand that
> this is not primarily a concern for Buckminster or b3, but maybe in this
> forum we can gather enough "weight" to convince the API Tools team that
> AB modularity is not less important than TP modularity. I wonder if they
> were willing to consider the reuse of the existing TP infrastructure
> instead of keeping their own, unmodular solution.
>
> Cheers
> /Eike
>
> ----
> http://thegordian.blogspot.com
> http://twitter.com/eikestepper
>
>
>
> Henrik Lindberg schrieb:
>> So, does this require new materializer, or is it the p2 based
>> materializer that does this?
>>
>> This leads to a different question - When materializing to a TP, or a
>> AB, I assume that the materialization itself does not make it the
>> active one (unless of course the materialization is performed to the
>> location of the active TP/AB).
>>
>> I can imagine specifying symbolic locations for "active TP", and
>> "active AB" - but this seems a bit dangerous.
>>
>> So, how can these settings be specified? In a query? What should
>> happen in UI mode? Maybe I don't want to switch at all (I am
>> populating my workspace using several queries, and I am going to
>> update AB etc.). Should it popup questions before switching? In
>> headless mode it should probably just switch if so directed?
>>
>> I also think this leads to issues of materializing/provisioning of
>> preferecens/settings.
>>
>> - henrik
>>
>>
>> Thomas Hallgren wrote:
>>>
>>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>>> have any bearing on b3 ?
>>>>
>>>> (Ability to set API baseline).
>>>>
>>> Yes, I think it will be relevant.
>>>
>>> An API baseline (AB for short) is very similar to a target platform.
>>> The difference is that a TP can be updated several times during a
>>> development cycle but the AB is often set right at the beginning of
>>> that cycle and then used as a verification mechanism to assert that
>>> nothing deviates from that AB. So the AB verifies both the workspace
>>> and the TP.
>>>
>>> Presently, the AB is simply a workspace global named entry that
>>> appoints a folder with plug-in's in it. Very similar to the TP but
>>> more limited in how it can be described. And like the current PDE TP,
>>> it's OSGi specific. I don't think it has to be though.
>>>
>>> A workspace can have many AB's but only one is active at any one
>>> time. The compiler uses both the AB and the TP while compiling.
>>>
>>> - thomas
|
|
|
Re: API baseline - relevant for b3 [message #488301 is a reply to message #488300] |
Sun, 27 September 2009 20:24 |
|
Thomas Hallgren schrieb:
> I agree with Ekie. I just don't understand why two different concepts
> exists that have that much overlap. I.e. Why isn't an AB just another
> TP? Or to rephrase it, why not define an AB by appointing a TP?
Maybe we should include one of the two PDE newsgroups in this discussion
to get their expertise? But which one? Or a bugzilla?
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
>
> - thomas
>
>
>
> On 09/27/2009 07:46 PM, Eike Stepper wrote:
>> Hi Henrik,
>>
>> You seem to be pointing to one of the weak areas of the API Tools
>> infrastructure: modularity of the API baseline (AB). I understand that
>> this is not primarily a concern for Buckminster or b3, but maybe in this
>> forum we can gather enough "weight" to convince the API Tools team that
>> AB modularity is not less important than TP modularity. I wonder if they
>> were willing to consider the reuse of the existing TP infrastructure
>> instead of keeping their own, unmodular solution.
>>
>> Cheers
>> /Eike
>>
>> ----
>> http://thegordian.blogspot.com
>> http://twitter.com/eikestepper
>>
>>
>>
>> Henrik Lindberg schrieb:
>>> So, does this require new materializer, or is it the p2 based
>>> materializer that does this?
>>>
>>> This leads to a different question - When materializing to a TP, or a
>>> AB, I assume that the materialization itself does not make it the
>>> active one (unless of course the materialization is performed to the
>>> location of the active TP/AB).
>>>
>>> I can imagine specifying symbolic locations for "active TP", and
>>> "active AB" - but this seems a bit dangerous.
>>>
>>> So, how can these settings be specified? In a query? What should
>>> happen in UI mode? Maybe I don't want to switch at all (I am
>>> populating my workspace using several queries, and I am going to
>>> update AB etc.). Should it popup questions before switching? In
>>> headless mode it should probably just switch if so directed?
>>>
>>> I also think this leads to issues of materializing/provisioning of
>>> preferecens/settings.
>>>
>>> - henrik
>>>
>>>
>>> Thomas Hallgren wrote:
>>>>
>>>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>>>> have any bearing on b3 ?
>>>>>
>>>>> (Ability to set API baseline).
>>>>>
>>>> Yes, I think it will be relevant.
>>>>
>>>> An API baseline (AB for short) is very similar to a target platform.
>>>> The difference is that a TP can be updated several times during a
>>>> development cycle but the AB is often set right at the beginning of
>>>> that cycle and then used as a verification mechanism to assert that
>>>> nothing deviates from that AB. So the AB verifies both the workspace
>>>> and the TP.
>>>>
>>>> Presently, the AB is simply a workspace global named entry that
>>>> appoints a folder with plug-in's in it. Very similar to the TP but
>>>> more limited in how it can be described. And like the current PDE TP,
>>>> it's OSGi specific. I don't think it has to be though.
>>>>
>>>> A workspace can have many AB's but only one is active at any one
>>>> time. The compiler uses both the AB and the TP while compiling.
>>>>
>>>> - thomas
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| | | |
Re: API baseline - relevant for b3 [message #578181 is a reply to message #488235] |
Sat, 26 September 2009 15:35 |
|
On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
> have any bearing on b3 ?
>
> (Ability to set API baseline).
>
Yes, I think it will be relevant.
An API baseline (AB for short) is very similar to a target platform. The difference is that a TP can be updated several
times during a development cycle but the AB is often set right at the beginning of that cycle and then used as a
verification mechanism to assert that nothing deviates from that AB. So the AB verifies both the workspace and the TP.
Presently, the AB is simply a workspace global named entry that appoints a folder with plug-in's in it. Very similar to
the TP but more limited in how it can be described. And like the current PDE TP, it's OSGi specific. I don't think it
has to be though.
A workspace can have many AB's but only one is active at any one time. The compiler uses both the AB and the TP while
compiling.
- thomas
|
|
|
Re: API baseline - relevant for b3 [message #578196 is a reply to message #488242] |
Sun, 27 September 2009 15:49 |
Henrik Lindberg Messages: 2509 Registered: July 2009 |
Senior Member |
|
|
So, does this require new materializer, or is it the p2 based
materializer that does this?
This leads to a different question - When materializing to a TP, or a
AB, I assume that the materialization itself does not make it the active
one (unless of course the materialization is performed to the location
of the active TP/AB).
I can imagine specifying symbolic locations for "active TP", and "active
AB" - but this seems a bit dangerous.
So, how can these settings be specified? In a query? What should happen
in UI mode? Maybe I don't want to switch at all (I am populating my
workspace using several queries, and I am going to update AB etc.).
Should it popup questions before switching? In headless mode it should
probably just switch if so directed?
I also think this leads to issues of materializing/provisioning of
preferecens/settings.
- henrik
Thomas Hallgren wrote:
>
> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>> have any bearing on b3 ?
>>
>> (Ability to set API baseline).
>>
> Yes, I think it will be relevant.
>
> An API baseline (AB for short) is very similar to a target platform. The
> difference is that a TP can be updated several times during a
> development cycle but the AB is often set right at the beginning of that
> cycle and then used as a verification mechanism to assert that nothing
> deviates from that AB. So the AB verifies both the workspace and the TP.
>
> Presently, the AB is simply a workspace global named entry that appoints
> a folder with plug-in's in it. Very similar to the TP but more limited
> in how it can be described. And like the current PDE TP, it's OSGi
> specific. I don't think it has to be though.
>
> A workspace can have many AB's but only one is active at any one time.
> The compiler uses both the AB and the TP while compiling.
>
> - thomas
|
|
|
Re: API baseline - relevant for b3 [message #578218 is a reply to message #488286] |
Sun, 27 September 2009 17:46 |
|
Hi Henrik,
You seem to be pointing to one of the weak areas of the API Tools
infrastructure: modularity of the API baseline (AB). I understand that
this is not primarily a concern for Buckminster or b3, but maybe in this
forum we can gather enough "weight" to convince the API Tools team that
AB modularity is not less important than TP modularity. I wonder if they
were willing to consider the reuse of the existing TP infrastructure
instead of keeping their own, unmodular solution.
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
Henrik Lindberg schrieb:
> So, does this require new materializer, or is it the p2 based
> materializer that does this?
>
> This leads to a different question - When materializing to a TP, or a
> AB, I assume that the materialization itself does not make it the
> active one (unless of course the materialization is performed to the
> location of the active TP/AB).
>
> I can imagine specifying symbolic locations for "active TP", and
> "active AB" - but this seems a bit dangerous.
>
> So, how can these settings be specified? In a query? What should
> happen in UI mode? Maybe I don't want to switch at all (I am
> populating my workspace using several queries, and I am going to
> update AB etc.). Should it popup questions before switching? In
> headless mode it should probably just switch if so directed?
>
> I also think this leads to issues of materializing/provisioning of
> preferecens/settings.
>
> - henrik
>
>
> Thomas Hallgren wrote:
>>
>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>> have any bearing on b3 ?
>>>
>>> (Ability to set API baseline).
>>>
>> Yes, I think it will be relevant.
>>
>> An API baseline (AB for short) is very similar to a target platform.
>> The difference is that a TP can be updated several times during a
>> development cycle but the AB is often set right at the beginning of
>> that cycle and then used as a verification mechanism to assert that
>> nothing deviates from that AB. So the AB verifies both the workspace
>> and the TP.
>>
>> Presently, the AB is simply a workspace global named entry that
>> appoints a folder with plug-in's in it. Very similar to the TP but
>> more limited in how it can be described. And like the current PDE TP,
>> it's OSGi specific. I don't think it has to be though.
>>
>> A workspace can have many AB's but only one is active at any one
>> time. The compiler uses both the AB and the TP while compiling.
>>
>> - thomas
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
Re: API baseline - relevant for b3 [message #578233 is a reply to message #488287] |
Sun, 27 September 2009 20:20 |
|
I agree with Ekie. I just don't understand why two different concepts exists that have that much overlap. I.e. Why isn't
an AB just another TP? Or to rephrase it, why not define an AB by appointing a TP?
- thomas
On 09/27/2009 07:46 PM, Eike Stepper wrote:
> Hi Henrik,
>
> You seem to be pointing to one of the weak areas of the API Tools
> infrastructure: modularity of the API baseline (AB). I understand that
> this is not primarily a concern for Buckminster or b3, but maybe in this
> forum we can gather enough "weight" to convince the API Tools team that
> AB modularity is not less important than TP modularity. I wonder if they
> were willing to consider the reuse of the existing TP infrastructure
> instead of keeping their own, unmodular solution.
>
> Cheers
> /Eike
>
> ----
> http://thegordian.blogspot.com
> http://twitter.com/eikestepper
>
>
>
> Henrik Lindberg schrieb:
>> So, does this require new materializer, or is it the p2 based
>> materializer that does this?
>>
>> This leads to a different question - When materializing to a TP, or a
>> AB, I assume that the materialization itself does not make it the
>> active one (unless of course the materialization is performed to the
>> location of the active TP/AB).
>>
>> I can imagine specifying symbolic locations for "active TP", and
>> "active AB" - but this seems a bit dangerous.
>>
>> So, how can these settings be specified? In a query? What should
>> happen in UI mode? Maybe I don't want to switch at all (I am
>> populating my workspace using several queries, and I am going to
>> update AB etc.). Should it popup questions before switching? In
>> headless mode it should probably just switch if so directed?
>>
>> I also think this leads to issues of materializing/provisioning of
>> preferecens/settings.
>>
>> - henrik
>>
>>
>> Thomas Hallgren wrote:
>>>
>>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>>> have any bearing on b3 ?
>>>>
>>>> (Ability to set API baseline).
>>>>
>>> Yes, I think it will be relevant.
>>>
>>> An API baseline (AB for short) is very similar to a target platform.
>>> The difference is that a TP can be updated several times during a
>>> development cycle but the AB is often set right at the beginning of
>>> that cycle and then used as a verification mechanism to assert that
>>> nothing deviates from that AB. So the AB verifies both the workspace
>>> and the TP.
>>>
>>> Presently, the AB is simply a workspace global named entry that
>>> appoints a folder with plug-in's in it. Very similar to the TP but
>>> more limited in how it can be described. And like the current PDE TP,
>>> it's OSGi specific. I don't think it has to be though.
>>>
>>> A workspace can have many AB's but only one is active at any one
>>> time. The compiler uses both the AB and the TP while compiling.
>>>
>>> - thomas
|
|
|
Re: API baseline - relevant for b3 [message #578258 is a reply to message #488300] |
Sun, 27 September 2009 20:24 |
|
Thomas Hallgren schrieb:
> I agree with Ekie. I just don't understand why two different concepts
> exists that have that much overlap. I.e. Why isn't an AB just another
> TP? Or to rephrase it, why not define an AB by appointing a TP?
Maybe we should include one of the two PDE newsgroups in this discussion
to get their expertise? But which one? Or a bugzilla?
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
>
> - thomas
>
>
>
> On 09/27/2009 07:46 PM, Eike Stepper wrote:
>> Hi Henrik,
>>
>> You seem to be pointing to one of the weak areas of the API Tools
>> infrastructure: modularity of the API baseline (AB). I understand that
>> this is not primarily a concern for Buckminster or b3, but maybe in this
>> forum we can gather enough "weight" to convince the API Tools team that
>> AB modularity is not less important than TP modularity. I wonder if they
>> were willing to consider the reuse of the existing TP infrastructure
>> instead of keeping their own, unmodular solution.
>>
>> Cheers
>> /Eike
>>
>> ----
>> http://thegordian.blogspot.com
>> http://twitter.com/eikestepper
>>
>>
>>
>> Henrik Lindberg schrieb:
>>> So, does this require new materializer, or is it the p2 based
>>> materializer that does this?
>>>
>>> This leads to a different question - When materializing to a TP, or a
>>> AB, I assume that the materialization itself does not make it the
>>> active one (unless of course the materialization is performed to the
>>> location of the active TP/AB).
>>>
>>> I can imagine specifying symbolic locations for "active TP", and
>>> "active AB" - but this seems a bit dangerous.
>>>
>>> So, how can these settings be specified? In a query? What should
>>> happen in UI mode? Maybe I don't want to switch at all (I am
>>> populating my workspace using several queries, and I am going to
>>> update AB etc.). Should it popup questions before switching? In
>>> headless mode it should probably just switch if so directed?
>>>
>>> I also think this leads to issues of materializing/provisioning of
>>> preferecens/settings.
>>>
>>> - henrik
>>>
>>>
>>> Thomas Hallgren wrote:
>>>>
>>>> On 09/26/2009 04:06 PM, Henrik Lindberg wrote:
>>>>> Does this issue https://bugs.eclipse.org/bugs/show_bug.cgi?id=290621
>>>>> have any bearing on b3 ?
>>>>>
>>>>> (Ability to set API baseline).
>>>>>
>>>> Yes, I think it will be relevant.
>>>>
>>>> An API baseline (AB for short) is very similar to a target platform.
>>>> The difference is that a TP can be updated several times during a
>>>> development cycle but the AB is often set right at the beginning of
>>>> that cycle and then used as a verification mechanism to assert that
>>>> nothing deviates from that AB. So the AB verifies both the workspace
>>>> and the TP.
>>>>
>>>> Presently, the AB is simply a workspace global named entry that
>>>> appoints a folder with plug-in's in it. Very similar to the TP but
>>>> more limited in how it can be described. And like the current PDE TP,
>>>> it's OSGi specific. I don't think it has to be though.
>>>>
>>>> A workspace can have many AB's but only one is active at any one
>>>> time. The compiler uses both the AB and the TP while compiling.
>>>>
>>>> - thomas
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| | | |
Goto Forum:
Current Time: Thu Apr 25 17:48:37 GMT 2024
Powered by FUDForum. Page generated in 0.05048 seconds
|