I love that Thom and Nigel continue to experiment in this space, and having been previously unfamiliar with Bundle I was pleasantly surprised to see that several of my favorite artists/labels have released material via this channel in recent months.
But I'm not sure if I really see the benefit of this (for artists or listeners) over e.g. Bandcamp, and the more I dig into BT's messaging about Bundle[0] the less I feel I understand about the overall strategy.
Elsewhere in the comments someone claims to have found his Facebook profile, and perhaps they did. But as a counterpoint, my hometown of Glendale (Northeast LA-adjacent) has the largest population of Armenians anywhere in the Western Hemisphere. Lots of Armenians == Lots of guys named Armon/Armen/Arman. We can safely assume at least two of them drive for Uber.
Just yesterday I was investigating music/production startups, and one of the few ideas I came across that seemed halfway-interesting was Splice[0]. On the producer side they have plugins for a few DAWs that enable easier collaboration and version tracking and so forth. But what comes of that (and why it's relevant to your comment) is a Soundcloud-ish embeddable web player that breaks out all the tracks/stems into a timeline beneath the main player interface.
It's pretty interesting, although I don't know how much of an opportunity is there, for some of the same reasons mentioned in the other replies. Musicians and producers can be incredibly reticent about this stuff.
The Soundplane from Madrona Labs[0] certainly merits a mention here. Pairs exceptionally well with their software synths Kaivo (physical modeling and granular synthesis) and Aalto (the closest thing to a Buchla in VST form).
I kind of assume at this point that most people who have an interest in this sort of thing know about monome[1], though I was surprised that there hasn't been more interest from HN in their most recent release, aleph[2], a (mostly) open source sound computer that can interface with just about anything that produces sound, cv, or bits.
Aleph is massively overpriced. For the same money you could get a laptop, a knobby controller, and a copy of Max/MSP or Reaktor, or Reason. Or a bunch of Eurorack modules of similar capabilities, or [fill in one of many blanks]. I think it would have to get under $700 to sell well; further more the tiny form factor is actually a negative because it doesn't give you a whole lot of hands-on control, nor is it suitable for performance ergonomically. The reaction I've heard from other electronic musicians has been essentially 'a solution in search of a problem.'
While Monome were among the first to innovate in this area, the steep price premium isn't justified when you can get similar or better functionality for less than half the price for most of their offerings.
On one hand, if you have no use for CV i/o aleph is basically a non-starter. Additionally, there are other CV solutions e.g. Expert Sleepers or the Kenton MIDI > CV boxes that get the job done a lot cheaper.
As you note, a substitute (of sorts) for anything they make can be had cheaper. Brian & Kelly et. al. have never been a concern oriented towards being the budget provider of anything, as they readily admit:
"monome is operated on a human scale. we use local suppliers and manufacturers with whom we've created long-term, trusting relationships. environmental and economic sustainability are critical considerations in our design process. we believe in beautiful design and quality craftsmanship. editions are produced in short runs according to demand. staying small affords the flexibility to pursue interesting new directions, not simply commodify established trends."
There is something to be said for the blank slate, raw unadulterated OSC, it only does exactly what you tell it to do design ethos. When I bought a monome I had never seen or used anything like it before. Though it should be noted that I sold my monome and kept my Launchpad, at least in part because the monome could be sold for exactly what I bought it for, and I'd be lucky to get $100 for Novation's hunk of molded plastic.
However, $1400 doesn't get you very far in Euro (especially if we're including the aforementioned ES modules to emulate aleph's USB>CV capabilities). One sound source, an EG and a filter? Maybe an additional source of modulation if you budget well? Not to mention power, rails and cases, cables and other accessories, etc. Ultimately, I defy anyone to find anyone with any experience in the matter who doesn't think modular synths are a money pit.
For that matter, Reaktor is basically just a springboard to spend more money on more Reaktor ensembles—largely because there are some teams doing really great stuff for that platform—and Max, plus an audio interface, plus a cheap plastic box of knobs and faders leaves you maybe $700 to buy a computer? It's not as if aleph is way out of line with the alternatives on price.
Where it does stand alone, and the reason I find it compelling, is that aleph is the first all-in-one box I'm aware of that offers the sound/data processing and marshaling capabilities of a computer without the "dude checking his email on stage" aspects of a laptop. That's something that I believe people have been awaiting anxiously, and although it's far from perfect, it's a start.
It's been really interesting over the last 24 hours or so to watch the reaction from Joe Kay[0], whose label/collective Soulection[1] has shared hundreds of hours of mixes and dozens of original tracks and remixes on Soundcloud[2], and to an extent (at least initially) built their following on the site (not to take anything away from the impact performances, etc. have had on their growing popularity), but were always at high risk of this sort of takedown due to remixes of popular rap/r&b tracks.
One of my initial takeaways from this development is that while Soulection may not have come to exist without Soundcloud, the "next Soulection" will almost certainly take root anywhere but Soundcloud.
And even if hip hop/trap/neo-soul/etc aren't your style, if you're into any electronic music at all there's probably something for you on Rinse.fm (unfortunately they're more an internet radio station than a platform like Soundcloud, not exactly apples to apples from a user perspective).
Rinse has an FM license now, so they've gone from pirate to legit. It was funny listening to the DJs trying to curb their swearing (50 pound fine per naughty word) in the weeks after they were licensed.
I submitted a two point critique via the Uservoice widget, one was merely a UI issue, another I feel might be worthy of a wider discussion.
I was excited to see this posted, and hoped it would be what I had been looking for. My friends and I had not yet found a suitable World Cup prediction competition site.
I immediately signed up, and was ready to send off some invites, when I found that Scoragora wanted me to predict a score for each of the 48 matches in group play, a fairly tedious exercise. Further, upon completion, it is completely unclear whether the game will be calculating winners and runners-up from the groups automatically, whether there will be an opportunity to pick winners in the knockout rounds—basically everything after group play is a mystery.
Suffice to say this is not what I expected. I can understand some need for score predictions—most likely in the later rounds for tie-breaking purposes (perhaps the best way to handle this is to ask users to predict the cumulative goal differential of the winning side). But score predictions ultimately aren't the core functionality of a game like this. If I predict that Brazil will defeat Mexico 4-0, and my friend predicts that Brazil will win 4-1, I don't really care who was right about the score, we both rightly acknowledged that Mexico is a shambles at the moment (poor Chicharito—he's only 26!) and that Brazil ought to win.
The functionality I expect to see front-and-center is, simply put, predictions about who wins each match through the final. I truly can't tell whether that functionality exists in Scoragora right now, and I don't really care to predict 96 scores for the privilege of finding out. Moreover, I'm the geek in my group of friends, there's at least some chance I might fill out the entire prediction form to satiate my curiosity about how you've constructed this web app, but I'm certain none of my friends will have the patience to do so, so I won't be inviting them, and we'll conduct our competition elsewhere.
UPDATE: I've just tried out Yahoo's game, and theirs is structured almost identically (though a bit prettier), so maybe my expectations are out of line? I don't understand why the game basically resets when the knockout rounds begin. If i predict that Spain and Chile will get out of their group, and a friend predicts Spain and the Netherlands, and another predicts the Netherlands and Chile, there should be some benefit for the one made the correct prediction and some penalty for the two who didn't.
Hi Ryan,
Thanks for the uservoice tickets (btw, uservoice is really awesome). This is great feedback thanks a lot.
* I'll definitely provide more info on what happens after group play. We say in the rules that "Predictions can be submitted at any time before the official start time of the concerned game", and this is true for all the games, including those in the knockouts. We thought that the knockout rounds would be extremely boring if your predictions are already set with the wrong teams. And you already got penalized for making the bad predictions in the group play. So we took the decision to open the knockout for predictions only when the teams are known, as it seemed to us that it would preserve the fun until very late in the competition.
* regarding the score inputs, actually I think this really depends on people. with my friends we wanted this level of details because it makes it more fun to us. To use your example, if I predict 4-0, you predict 4-1, and the actual score is 4-1, you will earn a few more points than me.
"The ruling confirms the need to bring today's data
protection rules from the 'digital stone age' into
today's modern computing world where data is no longer
stored on 'a server', or once launched online disappears
in cyberspace"
— European Commission vice-president Viviane Reding ("who has led the EU's data privacy efforts")
What—and I'm asking in earnest—is that quote even supposed to mean?
It just sounds like "once you publish something online it will be replicated in so many places that trying to get it removed is meaningless". Which is true and irrelevant to the question if people should be able to get things taken down from Google, a highly visible place being the leading search engine by a huge margin. It's like saying "because we can't erase every single physical record in the world you can't demand to take down that libelous billboard in Times Square". Not that I think Google is that billboard, it's just indexing what's already there.
Does this mean we should accept and ignore character smears, witness protection, cyberbullying and libel? Is it okay for someone to surrender the arbitrary association of content to their name which may not be true?
The solution might not be beautiful or scalable but removing search results could seriously improve some people's lives'.
I'm not talking about allowing people to remove information pertaining to trying to hide poor behaviour online or bad reviews online. I'm talking about the life ruining things.
Except this case was about a life-ruining thing that was true.
And raises some potential problems: suppose I default on my mortgage and get foreclosed, and you end up buying my house. Now, the fact that I was once the owner and that it changed hands in the foreclosure is part of the chain of ownership of that property, and may become necessary information if your ownership is ever challenged.
Does my "right to be forgotten" preclude your right to prove your ownership of the property?
It was a 1998 foreclosure. Under US law, items older than 7 years are stricken from credit reports (though other actions may remain part of a record). In a world in which arbitrary reasons for denial exist, being prejudiced by a 16 year old, or 26 year old, or 46 year old financial mishap would seem ... less than just.
How about the roughly 485,000 households in the US receiving foreclosure notices in the 2010 foreclosure crisis. I'm sorry, that was in September 2012 alone. If that's a persistent Web record, should that follow them around for the rest of their lives?
Or, in a different context, does a rhinoceros have a right to be forgotten? http://redd.it/25ll1v
There are several things your over-simplified analysis is missing here.
One is that the "right to be forgotten" does not exist in a vacuum -- it has to coexist with a bunch of other rights, and weighing how important each of them is in relation to the others is incredibly difficult. And that's without getting into as-yet-uncodified rights (example: making history more accessible via scanning/OCRing of old newspapers, which now seems like it'd have to come with a censorship regime built in to expunge news that's "meant to be forgotten").
Another is the practicality issue, in that information of this sort is actually incredibly hard to destroy, and often is required to be published in order to provide a fair process to everyone involved.
Finally, there's the issue that this is going after Google, but that doesn't actually accomplish the kind of "forgetting" you seem to be talking about. Sure, the general public won't see a newspaper listing about the foreclosure in Google anymore, but the people in a position to use that information to wreak institutional harm still will have access to it, because the databases they use are not publicly accessible, meaning it's not possible to determine if you're even in them in order to sue for removal.
Digitizing old newspapers isn't all that great an intrusion -- though it makes old content available, you're still bound by the bandwidth of those original print sources. It means that you might see what had been talk-of-the-small-town splashed around on national or international coverage now. In a levels-of-harm basis, it's not so bad.
On the weighted importance, one of the interesting elements I've seen said of Xeer (mentioned recently on HN: https://news.ycombinator.com/item?id=7736841) is that there's a concept of increasing obligations with increasing power or wealth (this is hearsay from a reddit comment, the Wikipedia article doesn't touch on this). It's similar to how I feel disclosure rules should operate. Someone with little or no impact on society should fear little disclosure. Someone with a great deal of impact (financial, political, military, religious, cultural power, or with a record of criminal acts causing suffering or death to others) should be obliged to disclose more of themselves.
As for coming up with automatic rules, I doubt that's possible, but then, it's not in law either -- that's why we've got judges.
Your criticism in going after Google has merits, though any law such as the EU one should address not only the access and indexing of such information, but of its use. Though proving someone knows a certain fact or accessed a specific piece of information is at best difficult.
That is a contrived example. In your example, the information would be kept, the fact the property was foreclosed would probably be stored in a cabinet or computer system in an office somewhere. You would make an effort to seek out and obtain that information legally and with justifiable cause. (To prove your ownership.) It would not be publicly accessible for the world to see for there is no reason for it to be. It would be a bit like medical records.
The core part of the argument to the right to be forgotten is to treat some pieces of information as medical information, i.e, private or not your concern. Things that you would rather not be known for they bestow little benefit to you except for you to profit at my expense. A lot of people find this uncomfortable, for good reason.
> The core part of the argument to the right to be forgotten is to treat some pieces of information as medical information
That sounds great, but this case clearly shows that's not the case at all. This is information that can be widely and freely reported, just not by everyone. A newspaper could run it as their front page story every day, but some other people can't report it, and, in fact, can't even tell you the fact that someone else is saying something related to the topic you're searching for. That's something very different than treating it as private data.
One thing about restricting public disclosures to established news organizations, as had been the case prior to the advent of the Web, is that there's a distinctly limited bandwidth for this. While _an_ individual could be smeared, the ability to do so on a mass basis was distinctly limited.
I haven't read the EU decision yet, nor have I made up my mind as to whether or not it does or doesn't have merits. In general I subscribe to the principle that there's a proportionality appropriate to disclosure: that the greater a person's (or institution's) power and responsibility, the greater the obligation for disclosure of relevant details of their life, most particularly as it might affect others (individuals, organizations, government(s), etc.).
She's saying that, in her view, people should own their information. Where Google (or anyone else) chooses to store it, should not be an excuse for denying what she sees as the right to manage that information.
At a high level, I imagine she would support a law that requires all information that can be associated with an individual (google account, ip address, etc) to be stored with metadata tracing it back to that identity and require that companies be able to remove all the data by metadata.
What do people find confusing about the quote? It seems like pretty standard politician to me.
> At a high level, I imagine she would support a law that requires all information that can be associated with an individual (google account, ip address, etc) to be stored with metadata tracing it back to that identity and require that companies be able to remove all the data by metadata
What does that have to do with linking to a factually true news story about a guy? It seems a lot more like she's trying to conflate things for a soundbite.
In any case, making all statements that are about information about someone transitively "personal information" seems like an awfully bad idea. That's great when Google can track an account or an IP address, but where's the metadata for Mario Gonzalez to be found in a newspaper's story about Mario Gonzalez? And considering that news stories are often about more than one thing, what's the balance when a story is about someone who wants the event to be forgotten and someone who doesn't (or it's important that it's not)?
In all seriousness though, in the UK credit record agencies can only keep 6 years of financial information - a search engine is not subject to any regulation around this and could for example return an insolvency event from 16 years ago.
Reding is saying this is a double standard and infringes on an EU citizens rights and that they'd like to regulate this.
A search engine does not know how old the contents of a page are. They'd have to diff different versions of a page, which is not foolproof; for html it's particularly noisy compared to, say, code. Even when that strategy works, or when there's only one version of a page, it's only possible to determine when the search engine first saw a piece of content, not when that piece of content was created.
In contrast, things on a credit history are dated.
Yeah I would have though Doom would be very high. But the density of his lyrics perhaps stem more from allusions/references and humour than from the words themselves.
But I'm not sure if I really see the benefit of this (for artists or listeners) over e.g. Bandcamp, and the more I dig into BT's messaging about Bundle[0] the less I feel I understand about the overall strategy.
[0]http://blog.bittorrent.com/2014/06/16/bittorrent-bundle-hits...