[
https://issues.jboss.org/browse/WFCORE-873?page=com.atlassian.jira.plugin...
]
Brian Stansberry edited comment on WFCORE-873 at 8/20/15 11:46 AM:
-------------------------------------------------------------------
[~kabirkhan] Right; please don't change that.
It's a manifestation of the fundamental mistake we made in having 'asterisk'
mean two different things. So in r-r-d we can't readily differentiate between wanting
the description of a single resource type whose value is 'asterisk' versus a
wildcard request where we want descriptions of all resource types that match a pattern.
The latter produces a list as the response, consistent with what read-resource does.
If we had used, for example, '?' to indicate a wildcard request instead of
'asterisk' then it wouldn't be ambiguous.
{code}/subsystem=infinispan/cache-container=*/invalidation-cache=*/eviction=EVICTION:read-resource-description{code}
would produce a non-list response.
{code}/subsystem=infinispan/cache-container=?/invalidation-cache=?/eviction=EVICTION:read-resource-description{code}
would produce a list response with a single element.
Some time over a beer we should discuss a strategy for eventually making this not
ambiguous. But we'd have to introduce something new, while still supporting the old
way for a long time, and then in some far off major release drop support for the old way.
was (Author: brian.stansberry):
[~kabirkhan] Right; please don't change that.
It's a manifestation of the fundamental mistake we made in having '*' mean two
different things. So in r-r-d we can't readily differentiate between wanting the
description of a single resource type whose value is '*' versus a wildcard request
where we want descriptions of all resource types that match a pattern. The latter produces
a list as the response, consistent with what read-resource does.
If we had used, for example, '?' to indicate a wildcard request instead of
'*' then it wouldn't be ambiguous.
{code}/subsystem=infinispan/cache-container=*/invalidation-cache=*/eviction=EVICTION:read-resource-description{code}
would produce a non-list response.
{code}/subsystem=infinispan/cache-container=?/invalidation-cache=?/eviction=EVICTION:read-resource-description{code}
would produce a list response with a single element.
Some time over a beer we should discuss a strategy for eventually making this not
ambiguous. But we'd have to introduce something new, while still supporting the old
way for a long time, and then in some far off major release drop support for the old way.
Resource alias operation inconsistencies
----------------------------------------
Key: WFCORE-873
URL:
https://issues.jboss.org/browse/WFCORE-873
Project: WildFly Core
Issue Type: Bug
Components: Domain Management
Reporter: Heiko Braun
Assignee: Kabir Khan
Priority: Critical
Fix For: 2.0.0.Beta3
Formlery we could request the ../eviction=EVICTION resource. I assume that with the
resource name change a transformer kicks that redirects the request to the
../component" => "eviction resource. The response header however contains the
wrong address, which prevents clients to correlate the request with the response.
*In HAL this breaks the access control mechanism.*
{noformat}
[domain@localhost:9990 /]
/profile=full/subsystem=infinispan/cache-container=*/invalidation-cache=*/eviction=EVICTION:read-resource-description
{
"outcome" => "success",
"result" => [{
"address" => [
("profile" => "full"),
("subsystem" => "infinispan"),
("cache-container" => "*"),
("invalidation-cache" => "*"),
("component" => "eviction")
],
"outcome" => "success",
[...]
{noformat}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)