Meridian Unplugged - Powered by MrTechGuy.co.uk

Lots of good ideas here, but I believe the system would benefit from a more fundamental conceptual shift.

The Sooloos design is consistent and rigorous � to a fault.

The conceptual model of the data, which of course drives the entire experience, is based on the formal thinking of relational databases. This rigor turns us all into librarians and curators. I have to do meticulous metadata editing when importing music, which is boring; but worse, I have to consider the rigorous data model when listening, when I just want to enjoy the music with serendipitous discoveries.

Why would rigor be bad? Consider an issue we keep discussing in these pages, whether to archive classical music by composer or conductor or orchestra or soloist. Many different styles are suggested, each defensible, none satisfying. If I browse my Beethoven albums, I might miss some filed under Bernstein or Grimaud. You might say that the fault is my inconsistent library, and my first answer is that I don�t want to be a meticulous librarian. But more fundamentally, I don�t like the rigor that insists on a single classification model. Nothing is only one thing.

This is an old-fashioned model. Modern systems focus on free-form search and link navigation in an informal relationship graph (tech jargon for clicking on the drummer in the credits).

In the Sooloos database, the primary index is a tree structure of Artists and Album. And there are secondary indexes for Credits, Genre, Tags, and all the rest. And these schema distinctions are reflected in the user experience. If I want to listen to Miles Davis, I can browse by Artist. But if I want to see other albums where Miles contributed as a player or composer, I have to use an entirely different technique, Focus by Credits � I can�t browse by Credits. And if I want to find tributary albums, like Bob Belden�s Miles Espa�ol, I need to use yet another technique, Search � but Search looks only at Artists and Albums and Tracks, it doesn�t look at Credits, so it wouldn�t find compositions by Miles.

And if I have a defined a Tag for Miles Davis style stuff, or want to look at a Genre like Bebop, how do I find that? Tags and Genres are only under Focus, I can�t browse Tags or Genres. Turns out I can both browse and Focus by Release date, but not Search.

Similarly, a Search for �Blues� finds Bluesbreakers and Paul Butterfield Blues Band, and a lot of tracks in the Jazz genre that use the word Blues in the title. But perversely, it doesn't find the albums in the Blues genre � search doesn�t look at Genre.

And in none of these listings do my Blues playlists show up! What�s up with that? A playlist is a very personal grouping of music. Groupings are useful. A Genre is an industry grouping; an Album is a publisher grouping (Telarc was pleased to group Rite of Spring with Nielsen�s 5th Symphony on an album with Paavo J�rvi); a Tag is a personal grouping; but the Playlist is personal and more intimate. But Playlists do not work for browsing or searching or focusing.

This is all random. Rigorous and consistent, but random.

These different types of navigation use different user interfaces, and different ways of canceling out. But my complaint is not with the inconsistency, it is with the formality. I do think some of the user interfaces for Focus and Search are poor, but that is not the point: those issues could be fixed. The fundamental issue is that the systems insists on the different user interfaces, it insists on the conceptual differences, on rigor.

There is a surfeit of concepts.

I am very focused on ad hoc browsing, on serendipitous discovery. This is what Sooloos promises, and still does better than others. Search is not the problem: if I know what I want I can certainly find it, in Sooloos and competitive systems. But that requires that the whole database is in my brain, and I browse and focus and search there, and then I use the system just to locate what I want. That is not what Sooloos is about. Sooloos should help me find interesting stuff I hadn�t thought of, inspire me, take me down new alleys, help me discover and learn. That must be the differentiation.

This rigorous approach is on the wrong side of history. Think of Google. Think that is easier to find an album to buy at Amazon than to find an album to play on my own system. Think that in Windows 8 (or similarly in iOS Spotlight) I just start typing Miles, and a list of possible candidate search teams come up; I select Miles Davis (as opposed to my coworker Miles Brown), and I see a list of my own Miles Davis albums, and the Miles Davis albums in the cloud that Windows can stream for me, and the Wikipedia and IMDB pages, and Amazon and HDtracks, and the official Miles Davis site, and photos�

Netflix handles this well: select a movie and it lists genres and actors and directors, and each item is clickable, in effect focusing on that actor or genre. It shouldn�t even be necessary to classify them as Genres and Actors -- I�m not confused about whether Al Pacino is an Actor or a Genre. And crucially, it doesn�t matter if Pacino is an Actor or a Genre: if I click on Al Pacino I want to see stuff with Al Pacino, regardless of how the database designer has chosen to use him. What if some curator decides that Al Pacino is such a cultural icon that he has become a Genre? Fine, I don�t care, I click on the name and I want to see what is relevant.

(Technically, we navigate a directed graph structure and the edges of the graph can be treated as untyped. They are strictly typed in the schema, the system does know the difference and we can allow strictly typed search for the archivists among us, but we don�t need to insist.)

The rigor is also the reason why we need to do so much editing of metadata on import. Albums come in labeled by composer or conductor or orchestra or soloist, and I have to clean it up, bringing them to a consistent model, because otherwise I can�t find them. If the system didn�t insist on these distinctions, I could just import them and accept the metadata, consistent or not, and navigate the same way.

It is important to have a test � have we succeeded? I suggest this: can we do without a manual, like the Sooloos guide for collections? Amazon, Netflix, HDtracks don�t have manuals.

WHAT IS THE SOLUTION?

How do we achieve that?

1. There is only one display format of albums, the main browse page.

2. You can specify how this is sorted: by Artist or Release Date or Import Date, but also by Genre or Credit or Tag or Playlist. Anything.

3. You can specify how to constrain the contents (filter it) which is a combination of Focus and Search. It works across all metadata: Artist and Album and Credits and Genre and Tag and Playlist and�

4. When I change the focus, it doesn�t change the sort order. When I change the sort order, it doesn�t change the focus.

5. The main approach for Focus is that you can click on any metadata in any presentation. You see an artist or a Genre or a date or a Tag mentioned on a display somewhere, you can Focus on it by clicking. Sure, there should also be a separate user interface for sophisticated filters, but I think clicking on links will be the most common technique.

6. Search: if you have a keyboard you just start typing, otherwise you touch a search icon to bring out an onscreen keyboard. Two common conveniences in search: as you type, character by character, the system presents a list of common search terms, and at the same time it begins to reduce the contents of the main lists. Start typing �blues� and the drop down list shows Blues (Genre), Bluesbreakers (Artist), Subatomic Particle Homesick Blues (Album), Ahmad�s Blues (Track), Chicago Blues (Tag), Favorite Blues (Playlist), and at the same time the main window shows music with any one of those. Select Bluesbreakers from the list, and the browse page shows albums by Bluesbreakers and with Bluesbreakers in the credits.

So that�s the first step, changing navigation. Once you click on an album, we could treat it like we do today.

(Yes, we need a display format for tracks as well, under the same principles.)

CONVENIENT LINK NAVIGATION

We can make another change to make this thing more useful: once you have opened an album, the first page should show a compact and efficient summary of all kinds of relevant metadata, in addition to the tracks. The main artists, the genres, tags, etc. All of those are clickable, which becomes the driver to make relationship-based focus convenient. The goal should be that there is nothing on the home page that is not clickable.

How do we make room for all this? We remove as much �chrome� as possible, lines and boxes and stuff that doesn�t serve a purpose. (This is the prevailing design trend anyway, started by Windows 8 and now in iOS 7/8 and all over the web.) And we remove less important data, like the instrument for each credit, and track times and number of plays.

We can still have the specific and more complete metadata pages separately. But the main page is key. (Look at Netflix.)

How do we select the �relevant� metadata out of all the credits? Aha� More on this later.

WHAT IS REQUIRED TO MAKE THIS MODEL WORK?

First, the browse page is no longer a strict tree structure, an album will exist under many names in Credits, or under many Genres or Tags. It becomes a �polyarchy�, allowing diamond structures (multiple parents). This is ok. The world is full of diamonds.

Second, this form of navigation encourages filter/search/focus willy-nilly, so simply looking at an artist means you lose the previous Focus, which may have been meticulously constructed. The system needs a robust �Back� function, just like we have in the web browser. Netflix does this very well: even a search presents the results in a standard listing, and you can still go back. And the memory for Back should be long-term, not just over the current session (because a session is a technical artifact, not central to my experience): I should be able to go back to what I looked at yesterday. Why not? We have a server to store this stuff.

Third, how do we make browsing by Credits meaningful? We can�t have a browse display with tens of thousands of people. First, we don�t want the photographer and the writer of liner notes or the engineer, we need a built-in filter for relevance. Still too many. It is tempting to allow manual control of which musicians to elevate to browsable, but that is the wrong approach, it brings us back to curating. We could include only those people who also show up as Artists, that way Miles Davis as composer would show up because I have Miles Davis as an Artist. Useful, but it would miss a lot, like bassists who never play as leaders but are still interesting: it would be good if the bassist Marc Johnson shows up because he plays with a lot of important musicians, even if I don�t have an album by Marc Johnson. We can probably come up with some sort of relevance algorithm by looking at relationships in the library� We could make the system learn: every time I focus on a credit, the system remembers and includes that person in the browse list. Useful, but it wouldn�t inspire serendipitous discovery.

But all of these useful tricks are classical thinking. It isn�t about my computer and my collection�I live in the cloud. How can we leverage community knowledge? What if all Sooloos users� browse history was sent up to the cloud, and collated, and the full community knowledge was sent down to each of us to guide the browse list? What if we analyze the metadata of other music services, and Amazon and Wikipedia? Interesting�

OTHER IMPROVEMENTS
...independent of the �less rigor� story

When I click on an artist, don�t ask me if I want to Focus or Swim. In fact, don�t ask me anything, just do it. (I have a very annoying car, when I enter an address it asks me if I want directions or want to store the address in the address book. SHUT UP AND JUST GET ME THERE! Imagine a phone where you dial a number and it asks if you want to call it or store it. SHUT UP AND JUST CALL THE NUMBER! Or in Sooloos: I select Blues, SHUT UP AND FOCUS. Then you add a Swim button that always acts on the current focus, without further questions. This means two clicks to focus+swim, but that�s what we have today; and we would have only one click for Focus. General rule: every action takes one click.

With the buttons for Play Now, Play Next, Add to queue, I would add �Wipe the queue and play�: I do it very frequently and today it requires six clicks.

Support multi-select. Every platform has some form of multi-select operation, ctrl-click with a mouse or some touch gesture. Indeed, Sooloos uses this for editing. Make it universal. Quick and convenient creation of a playlist: focus on Miles Davis, touch-touch-touch on a few albums, and the �wipe and play� button. Today this is four clicks to clean out the queue, and then for each album select, play all tracks, add to queue, back to browse list. Sixteen clicks to play three albums, instead of four.

The album detail display: I would recommend a horizontally scrolling �magazine� layout instead of the fixed page layout. The current display is not really a multi-touch interface, it is a mouse-click interface with the tabs for different kinds of metadata. The horizontally scrolling page fits better with different size screens: it isn�t just the iPad anymore, we will have lots of touch-screen sizes of iOS and Windows and Android. Real estate is no longer a scarce commodity, allowing more generous listings: Albums by this artist, Albums with this artist, Albums with these artists, Albums with related artists�

And finally, use this same display for editing, so we can get rid of of the antique Control:PC and Control:Mac. If I drill into one album, I can select Edit and everything becomes editable. And if I select multiple albums and edit, only the relevant fields like Genre and Tag show up and are editable, tracks are not. Same navigation technique for everything. On every device.

ENGINEERING

How can we afford all this? There will be different code for the platforms, there is no common tooling that can create a top-notch experience. Ouch. But with all the data and all the intelligence and all the state stores on the server, the user interface layer is relatively lightweight. And it can�t be avoided. And today, we have four or five programs: iPad, iPhone, Control:15, Control:PC, Control:Mac. With this model, you get three: Windows, iOS, Android.