Hello list,
I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
On 02/20/2012 05:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
I think that, historically, we gave those different names because they contained a different combination of privileges from the 'User' roles. However, now that some roles contain privileges from multiple levels (Pool admin has pool and instance privileges, etc) the distinction is no longer so relevant.
To clarify, your intent here is not to substitute the i18n strings when processing seeds.rb, but to actually put the i18-ready strings in the database, is this correct? i.e. instead of "Pool Administrator" the string in the db would have 't("obj_types.pool") t("roles.administrator")' -- or something like that? That way we'll be properly translating roles for the user's locale. Also, although the current set of role names fits nicely with "<foo> User" and "<foo> Administrator", it's not safe to assume this will always be the case. However, if what you're doing is just putting a string to be globalized into the db, that would work for more specific one-off roles that didn't fit this scheme so well.
On the other hand, having a separate string for each role name might be better in that it avoids the problem of word order being different depending on locale --i.e. "obj-type role-type" may be the correct order in english but perhaps not in all languages. You'd avoid this problem if you used t("roles.names.provider_administrator"), etc.
Scott
From a user experience standpoint, you're fine here. If everything else is in "<object> user" format, those should be too, unless there's a technical reason (which would be more than a little weird).
I think it's a bit odd, overall, to have the "user" suffix appended to every role name -- isn't it apparent in the UI which user you're connecting with which role, or must there be redundancy?
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On Feb 20, 2012, at 2:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
On 02/21/2012 11:04 AM, Kenneth Keiter wrote:
From a user experience standpoint, you're fine here. If everything else is in "<object> user" format, those should be too, unless there's a technical reason (which would be more than a little weird).
I think it's a bit odd, overall, to have the "user" suffix appended to every role name -- isn't it apparent in the UI which user you're connecting with which role, or must there be redundancy?
It's certainly not on every one -- in fact, 'User' is used as opposed to 'Owner' or 'Administrator' or some such. So in this case 'user' is not an example of redundancy -- it indicates a level of access less than that of an owner or admin or similar.
For example, a "Pool User" has access to a pool, can launch deployments, etc but can't edit the pool or see/modify anyone else's instances. A "Pool Owner", however, has full access to the pool -- edit pool properties, start/stop anyone's instances in the pool, etc. Also, it's not <object> user -- the role is "<object-type> user" (or "<object-type> owner", etc., but it's applied to a specific object. i.e. as a user, I may have the "Pool owner" role on "pool1" but "Pool user" role on "pool2" -- the role is defined as "Pool user" not "pool1 user" -- but it's _assigned_ to me on a specific pool -- but I may have different roles on different pools.
Scott
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On Feb 20, 2012, at 2:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
Ah, interesting. We should come up with a way to remove the requirement of having those in the name, in the future, so that we can group them in the UI by type.
Thanks Scott!
On Feb 21, 2012, at 8:29 AM, Scott Seago wrote:
On 02/21/2012 11:04 AM, Kenneth Keiter wrote:
From a user experience standpoint, you're fine here. If everything else is in "<object> user" format, those should be too, unless there's a technical reason (which would be more than a little weird).
I think it's a bit odd, overall, to have the "user" suffix appended to every role name -- isn't it apparent in the UI which user you're connecting with which role, or must there be redundancy?
It's certainly not on every one -- in fact, 'User' is used as opposed to 'Owner' or 'Administrator' or some such. So in this case 'user' is not an example of redundancy -- it indicates a level of access less than that of an owner or admin or similar.
For example, a "Pool User" has access to a pool, can launch deployments, etc but can't edit the pool or see/modify anyone else's instances. A "Pool Owner", however, has full access to the pool -- edit pool properties, start/stop anyone's instances in the pool, etc. Also, it's not <object> user -- the role is "<object-type> user" (or "<object-type> owner", etc., but it's applied to a specific object. i.e. as a user, I may have the "Pool owner" role on "pool1" but "Pool user" role on "pool2" -- the role is defined as "Pool user" not "pool1 user" -- but it's _assigned_ to me on a specific pool -- but I may have different roles on different pools.
Scott
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On Feb 20, 2012, at 2:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On 02/21/2012 11:33 AM, Kenneth Keiter wrote:
Ah, interesting. We should come up with a way to remove the requirement of having those in the name, in the future, so that we can group them in the UI by type.
Thanks Scott!
On Feb 21, 2012, at 8:29 AM, Scott Seago wrote:
It's not quite as simple as that. The roles are a collection of lower-level privileges and, for the most part, they fall into a smaller subset of "useful combinations (user, owner, etc) -- but especially as I know we're moving in the direction of even finer-grained permissions in the future, we won't always have such a predictable set. Also, in the future we'll have a role management UI (for now there is no UI to manipulate roles) which will allow admins to create/modify roles which define new combinations of allowed actions on a given resource (or at a global level).
That said, I'm sure there's a way to organize this that improves usability, but I don't think it's as simple as assuming the current role names are all we'll need to deal with.
Scott
On 02/21/2012 11:04 AM, Kenneth Keiter wrote:
From a user experience standpoint, you're fine here. If everything else is in "<object> user" format, those should be too, unless there's a technical reason (which would be more than a little weird).
I think it's a bit odd, overall, to have the "user" suffix appended to every role name -- isn't it apparent in the UI which user you're connecting with which role, or must there be redundancy?
It's certainly not on every one -- in fact, 'User' is used as opposed to 'Owner' or 'Administrator' or some such. So in this case 'user' is not an example of redundancy -- it indicates a level of access less than that of an owner or admin or similar.
For example, a "Pool User" has access to a pool, can launch deployments, etc but can't edit the pool or see/modify anyone else's instances. A "Pool Owner", however, has full access to the pool -- edit pool properties, start/stop anyone's instances in the pool, etc. Also, it's not<object> user -- the role is "<object-type> user" (or "<object-type> owner", etc., but it's applied to a specific object. i.e. as a user, I may have the "Pool owner" role on "pool1" but "Pool user" role on "pool2" -- the role is defined as "Pool user" not "pool1 user" -- but it's _assigned_ to me on a specific pool -- but I may have different roles on different pools.
Scott
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On Feb 20, 2012, at 2:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
I'm going to put that on my list of things to take a look at at some point in the future. It strikes me that it's one of those UX elements which might be confusing to some users. The current structure is dictated by PM requirements, right?
Thanks for the quick reply!
On Feb 21, 2012, at 8:38 AM, Scott Seago wrote:
On 02/21/2012 11:33 AM, Kenneth Keiter wrote:
Ah, interesting. We should come up with a way to remove the requirement of having those in the name, in the future, so that we can group them in the UI by type.
Thanks Scott!
On Feb 21, 2012, at 8:29 AM, Scott Seago wrote:
It's not quite as simple as that. The roles are a collection of lower-level privileges and, for the most part, they fall into a smaller subset of "useful combinations (user, owner, etc) -- but especially as I know we're moving in the direction of even finer-grained permissions in the future, we won't always have such a predictable set. Also, in the future we'll have a role management UI (for now there is no UI to manipulate roles) which will allow admins to create/modify roles which define new combinations of allowed actions on a given resource (or at a global level).
That said, I'm sure there's a way to organize this that improves usability, but I don't think it's as simple as assuming the current role names are all we'll need to deal with.
Scott
On 02/21/2012 11:04 AM, Kenneth Keiter wrote:
From a user experience standpoint, you're fine here. If everything else is in "<object> user" format, those should be too, unless there's a technical reason (which would be more than a little weird).
I think it's a bit odd, overall, to have the "user" suffix appended to every role name -- isn't it apparent in the UI which user you're connecting with which role, or must there be redundancy?
It's certainly not on every one -- in fact, 'User' is used as opposed to 'Owner' or 'Administrator' or some such. So in this case 'user' is not an example of redundancy -- it indicates a level of access less than that of an owner or admin or similar.
For example, a "Pool User" has access to a pool, can launch deployments, etc but can't edit the pool or see/modify anyone else's instances. A "Pool Owner", however, has full access to the pool -- edit pool properties, start/stop anyone's instances in the pool, etc. Also, it's not<object> user -- the role is "<object-type> user" (or "<object-type> owner", etc., but it's applied to a specific object. i.e. as a user, I may have the "Pool owner" role on "pool1" but "Pool user" role on "pool2" -- the role is defined as "Pool user" not "pool1 user" -- but it's _assigned_ to me on a specific pool -- but I may have different roles on different pools.
Scott
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
On Feb 20, 2012, at 2:50 PM, Jason Guiditta wrote:
Hello list, I was just looking at seeds.rb to see about extracting our role names into i18n files so we can leverage the configuration I added there last week, and noticed what seems to me to be a minor inconsistency. Disregarding role with the name 'Administrator', it seems most of our roles are named: "{object name} user" or "{object name} owner"
There are 2 exceptions I see to this - the Instance and Deployment objects. These both use 'Controller' instead of 'User'.
I would like to propose making those consistent as part of the patch I am working on, unless there is some reason they are like this of which I am unaware. Thoughts?
-j
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
-- Kenneth Keiter Sr. Software Engineer, Cloud BU kenkeiter@redhat.com http://twitter.com/kenkeiter IRC: tigger or tiggerer
aeolus-devel@lists.fedorahosted.org