roller-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ian Kallen <>
Subject Re: spaces in tags (from 3.1 discussion)
Date Fri, 24 Nov 2006 18:22:30 GMT
IMO, the spaces-as-tag-delimiter folks have it wrong since it forces 
hostile use of language; "radioastronomy" isn't a real compound word and 
it's hard on the eye.

The driving principle should be human usefulness to identify and 
organize what the post is about. In your case, you really would want 
both "astronomy" and "radio astronomy" ... there's not a lot of 
throughput on the latter (see 
but the stronger the co-occurrence the more related the tags are (which 
is good, btw); in other words there's a strong relationship between 
these two but not with "radio" and tag relationships help inferred 
topicality emerge.

Technorati will normalize "radio astronomy" to "radioastronomy" (but we 
don't do the reverse lexical analysis) at query time, however the former 
is more readable. We do this normalization as a convenience since it's 
rarely wrong but I don't think the compound words are people friendly. 
When folks tag with artificial compound words, underscores, plus signs, 
etc I don't think it adds to human readability (underscores are stripped 
in our queries and the plus signs are treated as url encoded). 
Naturally, links have to be URL encoded; so tagging your post this way 
seems quite natural and informative
<a href="" rel="tag">radio 
<a href="" rel="tag">astronomy</a>


Jeffrey Blattman wrote:
> maybe it's because i don't really "get" tagging, but what confuses me 
> is that some tags cannot be represented w/ a single word. for example, 
> i might want to tag something with "radio astronomy". tagging it with 
> "radio" or "astronomy" doesn't capture the intent at all. i would 
> naturally want/try to quote the phrase.
> p.s., you can use the "+" char to encode a space in a url ... a little 
> nicer than %2 or whatever.
> Allen Gilliland wrote:
>> My opinion is that the way it works now is still the best, where tags 
>> have to be a single word and if you want to do phrases then use 
>> underscores or dashes.
>> I think the main reason against allowing for tag phrases is 
>> complexity.  You are increasing the complexity both on the user input 
>> side as well as on the retrieval side and for what I consider a 
>> marginal benefit.  On the UI side of things I think it's confusing to 
>> many users to allow for quoted phrases.  Then on the retrieval side 
>> you are also confusing things because how do you get at the url for 
>> "modern art"? /tags/modern%2Bart ... that's not something users can 
>> type in by hand which is part of what's nice about forcing single 
>> word tags.
>> As for Elias' suggestion of allowing phrases and converting the 
>> spaces to underscores, I think that's a little dicey.  For one, it 
>> still leaves the user confusion aspect around using quoted phrases, 
>> which I believe most users don't really want.  On the technical side 
>> I think you may be playing with fire though because why should the 
>> tag "modern art" become modern_art instead of modern-art.  And 
>> regardless of which you convert to, is that the value stored in the 
>> db?  That would mean that when the user comes back to that entry the 
>> tags list will show modern_art instead of their original tag phrase 
>> which can be confusing for users as well.
>> My belief is that the reason why tagging has been successful at 
>> places like and flickr is because the rules are short and 
>> simple. Tags are separated by spaces, period.  It's a slight blow in 
>> functionality but keeps the usability as simple and easy as possible.
>> -- Allen
>> Anil Gangolli wrote:
>>> Oh.  I must have missed the discussion about not supporting spaces 
>>> in tags because I would at least have made an attempt to convince 
>>> people to support spaces in tags.
>>> It comes up for tags like "modern art" that are not really 
>>> meaningful to separate.
>>> Regarding Elias's proposal, I initially felt against it, because I'd 
>>> rather use real spaces in the tags.  That's still the case, but less 
>>> so, because I did a quick Technorati sampling, and there already 
>>> seems to be a sizable rift between the true space users and the 
>>> underscore users.
>>> So for example Technorati lists 494 posts using "modern_art" 
>>> compared to 981 using "modern art" and differences in the thousands 
>>> for "george bush" compared to "george_bush", with the latter tag far 
>>> in the lead.
>>> --a.
>>> ----- Original Message ----- From: "Elias Torres" <>
>>> To: <>
>>> Sent: Wednesday, November 22, 2006 4:09 PM
>>> Subject: Re: Apache Roller 3.1RC1 (incubating) ready for testing
>>>> Allen Gilliland wrote:
>>>>> Anil Gangolli wrote:
>>>>>> I installed 3.1RC1 on my dev box.  Seemed to be clean using a 
>>>>>> fresh db
>>>>>> installation and 3.0 required-jars package.
>>>>>> I was trying the tag-related functionality out and I see a few 
>>>>>> issues:
>>>>>> (1)  The tag entry field on the entry edit page uses space 
>>>>>> separation
>>>>>> and doesn't seem to accomodate tags that include spaces.  I tried
>>>>>> quoting with double-quotes; that didn't seem to work.
>>>>> Yes, that is the expected behavior.  We talked about this when we 
>>>>> were
>>>>> evaluating the tags proposal and decided that most sites seem to be
>>>>> using the solution that we have, where tags cannot be multi word 
>>>>> phrases.
>>>> I now have a requirement at IBM to support spaces with the following
>>>> caveat. If you enter "elias is cool" it will be stored as 
>>>> elias_is_cool.
>>>> In other words, we support it as an input (double quotes) but we don't
>>>> store it that way. Are you guys cool with it?
>>>> -Elias
>>>>>> (2) The user guide link in the JSP footer seems to point off to 
>>>>>> an old
>>>>>> 2.x guide.  I was looking for and couldn't find the documentation
>>>>>> the 3.1 guides on the various forms of URLs, specifically I was
>>>>>> looking to test the tag-based URLs.  I know I've seen it somewhere,
>>>>>> but I can't remember where.
>>>>> Hmm.  For the user guide I think it would be nice if these kinds of
>>>>> links pointed to urls within the app, like /roller-ui/docs/*, for 
>>>>> user
>>>>> documentation about the current app version.
>>>>> I may also take this opportunity to throw out an idea I had a little
>>>>> while ago for documentation that was related to this.  It seems to me
>>>>> that since blogging is supposed to website publishing made easy 
>>>>> then one
>>>>> of the key components of a really mature blog system would be good
>>>>> documentation throughout the application.  I haven't been a big
>>>>> documentation contributor in the past, but I think that's more 
>>>>> because I
>>>>> felt the app needed more work and less docs at the time.  Now that 
>>>>> the
>>>>> app is getting more and more mature it may be time to consider a nice
>>>>> solution for providing rich documentation.
>>>>> So what I had been thinking about was a way where we could write 
>>>>> all of
>>>>> our documentation in small and easily reusable components, 
>>>>> possibly in
>>>>> xml, which we could easily use to either 1) provide a full help guide
>>>>> document (aka user guide) or 2) take bits and pieces of the docs 
>>>>> and be
>>>>> able to insert them directly into the appropriate pages.
>>>>> So for example, if a user is on the 'Templates' page and is 
>>>>> working on
>>>>> customizing their blog then we could have some documentation hooks 
>>>>> which
>>>>> provide contextual help info like ...
>>>>> 1. what is this page for?
>>>>> 2. what can i do on this page?
>>>>> 3. how do i use this page?
>>>>> 4. what do each of the fields on this page mean?
>>>>> 5. how does this affect my blog?
>>>>> So for the 'Templates' page the top of the page may provide quick 
>>>>> links
>>>>> which give a couple paragraphs of text explaining what the page is 
>>>>> for
>>>>> and what you do on the page.  Then you have the large text area where
>>>>> you can modify your templates which could have a little tool tip icon
>>>>> next to it which would tell the user what that field is for.  Then
>>>>> possibly at the bottom of the page we include a quick reference 
>>>>> sheet of
>>>>> the models and macros to help users while they are authoring 
>>>>> templates.
>>>>> One of the things that I think Roller has been hurting on is 
>>>>> usability
>>>>> and in my mind of the most important elements of usability when it 
>>>>> comes
>>>>> to web tools is contextual documentation.  I think doing something 
>>>>> like
>>>>> this could really help make Roller a more user friendly blogging 
>>>>> system.
>>>>>> (3) Tags don't seem to be displayed anywhere in the "basic" theme
>>>>>> all. Shouldn't we update this to show tags on entries, including
>>>>>> actual rel tag ?
>>>>> I think there may not have been any real consensus about whether 
>>>>> or not
>>>>> to promote tags in the themes since in many cases users may not be 
>>>>> using
>>>>> tags.
>>>>>> (4) I couldn't get anything to show up in the "Hot Tags" area of
>>>>>> front page.  Haven't investigated what is happening yet; this may
>>>>>> my own problem.
>>>>> Dunno about that one.
>>>>> -- Allen
>>>>>> --a.
>>>>>> ----- Original Message ----- From: "Dave" <>
>>>>>> To: <>
>>>>>> Sent: Monday, November 20, 2006 5:57 PM
>>>>>> Subject: Apache Roller 3.1RC1 (incubating) ready for testing
>>>>>>> Thank you and yes, I meant 3.1.
>>>>>>> With releasing 3.1, re-releasing 2.3.1 and working on 3.2 I've

>>>>>>> got a
>>>>>>> couple
>>>>>>> too many versions floating around in my head.
>>>>>>> - Dave
>>>>>>> On 11/20/06, Jeffrey Blattman <>
>>>>>>>> did you mean 3.1 RC1, or are we skipping 3.1?
>>>>>>>> Dave wrote:
>>>>>>>> > I've merged all applicable bug fixes from trunk to the

>>>>>>>> roller_3.1
>>>>>>>> > branch and prepared a first release candidate for the
>>>>>>>> release.
>>>>>>>> You
>>>>>>>> > can find the release files and latest 3.1 docs here:
>>>>>>>> >
>>>>>>>> >
>>>>>>>> >
>>>>>>>> > Here's the What's New in Roller 3.1 page:
>>>>>>>> >   
>>>>>>>> >
>>>>>>>> > Release candidates are for testing purposes only.
>>>>>>>> >
>>>>>>>> > Please help out. The sooner you download, test and report

>>>>>>>> bugs the
>>>>>>>> > sooner we'll be able to fix them and get the release
out. So 
>>>>>>>> please
>>>>>>>> > help out the project and take RC1 for a spin.
>>>>>>>> >
>>>>>>>> > - Dave

Ian Kallen || Architect, Technorati Inc. || m: 415.505.5208

View raw message