You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@archiva.apache.org by Joakim Erdfelt <jo...@erdfelt.com> on 2007/10/23 05:17:22 UTC
MRM-549 & MRM-547 : Proxy Connector Policy settings.
There seems to be some confusion on the settings, defaults values,
meanings, purpose, etc...
http://jira.codehaus.org/browse/MRM-549
* MRM-549 : proxy connectors: no "always" option for releases and
snapshots policies
http://jira.codehaus.org/browse/MRM-547
* MRM-547 : proxy connectors: cache failures options are confusing
http://docs.codehaus.org/display/MAVENUSER/Archiva+Proxy+Policies
I'd like to hear from you about what is bad about the current settings?
What is good about the current settings?
Some options on how to correct this?
(my 2 bits)
1) Create a sidebar on the proxy connector screen detailing the
meaning of the policy settings.
2) Change the policy setting values to make more sense to the largest
body of individuals.
I'd like to get these closed out, it'll be a simple fix, but the
decision needs discussion first.
--
- Joakim Erdfelt
joakim@erdfelt.com
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Brett Porter <br...@apache.org>.
yep, just under checksums I'd append "if they were previously
incorrect" to the description for FIX and "on downloaded resources"
to IGNORE.
Thanks,
Brett
On 01/11/2007, at 11:55 AM, Joakim Erdfelt wrote:
> Brett Porter wrote:
>>> (old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
>>> (proposed) ALWAYS, ONCE, HOURLY, DAILY, NEVER
>>> Snapshots (for often to check for)
>> Releases (how often to check for)
>> AS ABOVE
>>> Cache-Failures
>>> (old) IGNORED, CACHED
>>> (new) NO, YES
>>> Checksum
>>> (old) IGNORED, FIX, FAIL
>>> (new) IGNORE, FIX, FAIL
> <redstripe>Hooray List!</redstripe>
>
> New descriptions.
>
> (for releases / snapshots policies)
> ALWAYS : Means that the artifact is always updated from the remote
> repo.
> ONCE : Means the artifact is updated only ever ONCE from the remote
> repo. If it exists on the local repo, the remote repo is never hit
> for that artifact.
> HOURLY : Means the artifact is updated from the remote repo, only
> if it is at least one hour old on the local repo.
> DAILY : Means the artifact is updated from the remote repo, only if
> it is at least 1 day old on the local repo.
> NEVER : Means the artifact is never updated from the remote repo.
>
> (for cache-failures)
> NO : Means that the existence of old failures is not checked. All
> resource requests are allowed thru to the remote repo.
> YES : Means that the existence of old failures is checked, and will
> prevent the request from being performed against the remote repo.
>
> (for checksum)
> IGNORE : Means the contents / validity of the checksum files is
> ignored.
> FIX : Means that the checksum files are regenerated from the
> downloaded resources.
> FAIL : Means that the checksum files are validated against the
> downloaded resource, if the resource does not pass the checksum
> validation, the resource and the checksum files are deleted from
> the local repository and a failure is issued for the transfer.
>
> - Joakim
>
--
Brett Porter - brett@apache.org
Blog: http://www.devzuz.org/blogs/bporter/
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Joakim Erdfelt <jo...@erdfelt.com>.
Brett Porter wrote:
>> (old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
>> (proposed) ALWAYS, ONCE, HOURLY, DAILY, NEVER
>> Snapshots (for often to check for)
> Releases (how often to check for)
> AS ABOVE
>> Cache-Failures
>> (old) IGNORED, CACHED
>> (new) NO, YES
>> Checksum
>> (old) IGNORED, FIX, FAIL
>> (new) IGNORE, FIX, FAIL
<redstripe>Hooray List!</redstripe>
New descriptions.
(for releases / snapshots policies)
ALWAYS : Means that the artifact is always updated from the remote repo.
ONCE : Means the artifact is updated only ever ONCE from the remote
repo. If it exists on the local repo, the remote repo is never hit for
that artifact.
HOURLY : Means the artifact is updated from the remote repo, only if it
is at least one hour old on the local repo.
DAILY : Means the artifact is updated from the remote repo, only if it
is at least 1 day old on the local repo.
NEVER : Means the artifact is never updated from the remote repo.
(for cache-failures)
NO : Means that the existence of old failures is not checked. All
resource requests are allowed thru to the remote repo.
YES : Means that the existence of old failures is checked, and will
prevent the request from being performed against the remote repo.
(for checksum)
IGNORE : Means the contents / validity of the checksum files is ignored.
FIX : Means that the checksum files are regenerated from the downloaded
resources.
FAIL : Means that the checksum files are validated against the
downloaded resource, if the resource does not pass the checksum
validation, the resource and the checksum files are deleted from the
local repository and a failure is issued for the transfer.
- Joakim
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Brett Porter <br...@apache.org>.
On 01/11/2007, at 11:23 AM, Joakim Erdfelt wrote:
> Making no common ground between the options is also confusing.
why? they are not common behaviour! Two of them are in the context of
time, one a choice based on correctness, and the other a boolean
behaviour.
> There needs to be a common "DO NOT USE" or "IGNORED" option.
In the code, but not when you are trying to explain what it does.
If you really wanted to treat these this way in the UI, the only
acceptable way I would represent it is as an enable/disable checkbox
separate to the list of choices so the contexts don't get mixed. But
I don't think it's necessary given the options below where you can
translate it from English into that option.
> But I can release that opinion in favor of a clear and defined set
> of values.
> So far, all I'm getting is handwaving, big picture stuff, not a
> defined, concrete, list to work off of.
I gave you all my answers, but let me paste it in your format to be
crystal clear.
> Releases (how often to check for)
> (old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
> (proposed) ALWAYS, ONCE, HOURLY, DAILY, NEVER
> Snapshots (for often to check for)
AS ABOVE
> Cache-Failures
> (old) IGNORED, CACHED
> (new) NO, YES
> Checksum
> (old) IGNORED, FIX, FAIL
> (new) IGNORE, FIX, FAIL
Ok?
- Brett
--
Brett Porter - brett@apache.org
Blog: http://www.devzuz.org/blogs/bporter/
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Joakim Erdfelt <jo...@erdfelt.com>.
Brett Porter wrote:
> On 01/11/2007, at 10:01 AM, Joakim Erdfelt wrote:
>
>> IGNORED is a universal setting that means ... "This policy is ignored."
>> It can be applied to any policy.
>>
>> Changing IGNORED to ALWAYS is makes no sense for the other policies.
>
> I wasn't suggesting this for anything but release and snapshots (it
> was an eg, not an ie). For cache: it's a yes/no answer, and for
> checksum ignore is valid.
>
>> What about using SKIP in place of IGNORED?
>> What about using REJECT in place of DISABLED?
>
> I'm still confused what that really means.
>
> In the first set, to me, SKIP means "don't ever check for updates",
> since the context is when to check for updates, right? But I think
> you're saying it means "skip the check about whether to check for
> updates and check for updates anyway".
>
> If the question to be asked is "how often should Archiva check for
> updates"? SKIP, DISABLED, IGNORED, REJECT are not valid answers in the
> context of what happens. ALWAYS and NEVER would be.
>
> Or am I asking the wrong question? If so, what is a question you can
> ask and have each option as an answer for that matches the current
> behaviour?
>
> Likewise:
>
> * Should I cache failures? YES or NO. Or it could be "What should I do
> when I encounter a failure?" CACHE or DON'T CACHE. But
> IGNORED/SKIP/DISABLED/REJECT don't make sense in this context (at
> least with the given behaviour)
>
> * What should I do when a checksum is invalid? IGNORE, FIX or FAIL.
>
> Make sense?
No.
Making no common ground between the options is also confusing.
There needs to be a common "DO NOT USE" or "IGNORED" option.
But I can release that opinion in favor of a clear and defined set of
values.
So far, all I'm getting is handwaving, big picture stuff, not a defined,
concrete, list to work off of.
As I intepret your last comments, I see the list now as the following ...
Releases
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(proposed) SKIP, ONCE, HOURLY, DAILY, REJECT
Snapshots
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(new) SKIP, ONCE, HOURLY, DAILY, REJECT
Cache-Failures
(old) IGNORED, CACHED
(new) NO, YES
Checksum
(old) IGNORED, FIX, FAIL
(new) SKIP, FIX, FAIL
Any more changes?
- Joakim
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Brett Porter <br...@apache.org>.
On 01/11/2007, at 10:01 AM, Joakim Erdfelt wrote:
> IGNORED is a universal setting that means ... "This policy is
> ignored."
> It can be applied to any policy.
>
> Changing IGNORED to ALWAYS is makes no sense for the other policies.
I wasn't suggesting this for anything but release and snapshots (it
was an eg, not an ie). For cache: it's a yes/no answer, and for
checksum ignore is valid.
> What about using SKIP in place of IGNORED?
> What about using REJECT in place of DISABLED?
I'm still confused what that really means.
In the first set, to me, SKIP means "don't ever check for updates",
since the context is when to check for updates, right? But I think
you're saying it means "skip the check about whether to check for
updates and check for updates anyway".
If the question to be asked is "how often should Archiva check for
updates"? SKIP, DISABLED, IGNORED, REJECT are not valid answers in
the context of what happens. ALWAYS and NEVER would be.
Or am I asking the wrong question? If so, what is a question you can
ask and have each option as an answer for that matches the current
behaviour?
Likewise:
* Should I cache failures? YES or NO. Or it could be "What should I
do when I encounter a failure?" CACHE or DON'T CACHE. But IGNORED/
SKIP/DISABLED/REJECT don't make sense in this context (at least with
the given behaviour)
* What should I do when a checksum is invalid? IGNORE, FIX or FAIL.
Make sense?
>
> Brett Porter wrote:
>> I'd say 2)
>>
>> please just change "ignored" to a value that interacts with the
>> artifact (eg, always) instead of the policy, since that's what all
>> the other values do.
>>
>>
--
Brett Porter - brett@apache.org
Blog: http://www.devzuz.org/blogs/bporter/
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Joakim Erdfelt <jo...@erdfelt.com>.
IGNORED is a universal setting that means ... "This policy is ignored."
It can be applied to any policy.
Changing IGNORED to ALWAYS is makes no sense for the other policies.
Releases
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(new) ALWAYS, ONCE, HOURLY, DAILY, DISABLED
Snapshots
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(new) ALWAYS, ONCE, HOURLY, DAILY, DISABLED
Cache-Failures
(old) IGNORED, CACHED
(new) ALWAYS, CACHED
Checksum
(old) IGNORED, FIX, FAIL
(new) ALWAYS, FIX, FAIL
Using the logic that we should use wording that applies the artifact,
(which itself is misleading, as these affect metadata too), then the
DISABLED option should be changed too.
I have a feeling that we are confusing the meanings of IGNORED and DISABLED.
What about using SKIP in place of IGNORED?
What about using REJECT in place of DISABLED?
So it would become ....
Releases
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(proposed) SKIP, ONCE, HOURLY, DAILY, REJECT
Snapshots
(old) IGNORED, ONCE, HOURLY, DAILY, DISABLED
(new) SKIP, ONCE, HOURLY, DAILY, REJECT
Cache-Failures
(old) IGNORED, CACHED
(new) SKIP, CACHED
Checksum
(old) IGNORED, FIX, FAIL
(new) SKIP, FIX, FAIL
WDYT?
- Joakim
Brett Porter wrote:
> I'd say 2)
>
> please just change "ignored" to a value that interacts with the
> artifact (eg, always) instead of the policy, since that's what all the
> other values do.
>
> On 23/10/2007, at 1:17 PM, Joakim Erdfelt wrote:
>
>> There seems to be some confusion on the settings, defaults values,
>> meanings, purpose, etc...
>>
>> http://jira.codehaus.org/browse/MRM-549
>> * MRM-549 : proxy connectors: no "always" option for releases and
>> snapshots policies
>>
>> http://jira.codehaus.org/browse/MRM-547
>> * MRM-547 : proxy connectors: cache failures options are confusing
>>
>> http://docs.codehaus.org/display/MAVENUSER/Archiva+Proxy+Policies
>>
>> I'd like to hear from you about what is bad about the current settings?
>>
>> What is good about the current settings?
>>
>> Some options on how to correct this?
>> (my 2 bits)
>> 1) Create a sidebar on the proxy connector screen detailing the
>> meaning of the policy settings.
>> 2) Change the policy setting values to make more sense to the
>> largest body of individuals.
>>
>> I'd like to get these closed out, it'll be a simple fix, but the
>> decision needs discussion first.
>>
>> --
>> - Joakim Erdfelt
>> joakim@erdfelt.com
>>
>
> --
> Brett Porter - brett@apache.org
> Blog: http://www.devzuz.org/blogs/bporter/
Re: MRM-549 & MRM-547 : Proxy Connector Policy settings.
Posted by Brett Porter <br...@apache.org>.
I'd say 2)
please just change "ignored" to a value that interacts with the
artifact (eg, always) instead of the policy, since that's what all
the other values do.
On 23/10/2007, at 1:17 PM, Joakim Erdfelt wrote:
> There seems to be some confusion on the settings, defaults values,
> meanings, purpose, etc...
>
> http://jira.codehaus.org/browse/MRM-549
> * MRM-549 : proxy connectors: no "always" option for releases and
> snapshots policies
>
> http://jira.codehaus.org/browse/MRM-547
> * MRM-547 : proxy connectors: cache failures options are confusing
>
> http://docs.codehaus.org/display/MAVENUSER/Archiva+Proxy+Policies
>
> I'd like to hear from you about what is bad about the current
> settings?
>
> What is good about the current settings?
>
> Some options on how to correct this?
> (my 2 bits)
> 1) Create a sidebar on the proxy connector screen detailing the
> meaning of the policy settings.
> 2) Change the policy setting values to make more sense to the
> largest body of individuals.
>
> I'd like to get these closed out, it'll be a simple fix, but the
> decision needs discussion first.
>
> --
> - Joakim Erdfelt
> joakim@erdfelt.com
>
--
Brett Porter - brett@apache.org
Blog: http://www.devzuz.org/blogs/bporter/