almost 8 years on 2009-10-19


call me a child of an earlier generation... but i still get a kick out of looking at analytics and seeing that every day users in almost every country on earth use (ok, looks like we are missing about 5) - there is something really gratifying about that... only in the internet world can 12 people sitting in brooklyn provide services to people across the earth.

over the last few months, more and more users in various countries have been offering to translate our site for us. generally what they do is take a screenshot of a page or two, and then overlay the chinese or french words for each function. we really appreciate the gesture, but the sad reality is that getting a website properly translated to function in other languages is a really hard problem. our options have been:

#1 don't do anything on our end, assume that english is the lingua franca of the web, and let the translations happen at the edge (, etc.) - this is a terribly inefficient way to approach the issue, and leaves us with little control/ability to provide good accurate translations of our service. the babelfish just isn't real yet.

#2 invest a ton of resources in prepping our service for translation: this requires significant overhead for a growing service - generally it means turning each static text field into a table, where you can slot in each language. this isn't just a one-time cost, it is an ongoing drag on the speed of development.

#2.a. then spend a ton of money translating each language we want to use: we would have to work with a consulting firm, or some other agency for translation. from our quick research these guys are very expensive and not designed for growing startups

#2.b. set up a system to crowd source translation: works for facebook, but takes massive massive scale to do and takes a ton of curation anyway. not a practical option.

or - put differently - there is significant one-time and ongoing technology cost to translation, and then after that there is a significant human resource/organizational cost to building up and maintaining languages manually. so, as much as we would like to be in all languages, it just isn't realistic using traditional means.

enter 'the cloud' & service by

after i had been complaining about this issue for a while, jack welde, a new york entrepreneur, proposed a very modern solution -- why not build a company that exclusively focuses on the tech infrastructure and translation sourcing to allow companies like to just 'turn on' translation with a few html tags. the idea is to let a third party build and maintain custom human translated text tables off-site, and do real-time delivery of translated content. basically, why not use a little bit of dns magic, and allow a third party service to build up and maintain all the translations for a site like -- so, if someone requests a page in spanish, a third party service delivers all of the static text in the correct language via its translation delivery network, similar to the way that akamai delivers static assets from a server near your physical location. to over simplify, this is exactly how google translate works, but rather than relying on google and the users on the edge - with jack's company each service can get and control phrase specific high quality human translations, and jack's service can deliver translated interfaces in a matter of a few milliseconds...

why does this work? a big appeal to us is that we are both using amazon web services, so the bandwidth between our services is pretty close to free. basically, for this to work, needs to send smartling a request (not the assets), - smartling then needs to do on the fly translation delivery (a few miliseconds), and deliver the relevant interface translations to a user on the edge. since both and smartling use amazon web services, the transit between our services is in the same network, and is very very fast and low cost. earlier i have written a lot about how the value of 'clouds' actually compounds. amazon web services is now more valuable to because smartling is in there... our services can reliably talk to each other very quickly, and for incredibly low cost.

and, of course, this fits completely with a verticalizing vision of the internet. facilitates content and rich media between exactly whom you want how you want - across platforms and in real-time. we are not a translation company, just as we are not a server rack maintenance company, a distribution company, an identity provider, or a restaurant. companies are getting better and better at doing one thing within an ecosystem of quickly and cheaply interoperable systems. we are happy to not do our own translation. of course, this verticalization comes at the risk i have outlined several times of apis on apis on apis looking a lot like packaged financial products, but with proper research and a measured approach with a good backstop, we are happy that samrtling is providing us with a lot of good, clean leverage.

so, check out as an example of where the world is going - and check out in spanish and chinese - and let us know what you think.

p.s. sf - take notice, us little nyc startups are starting to breed and multiply... ;)

p.p.s ftc, in case it isn't totally clear, does have an economic arrangement with these guys

original swl blogposts and letters 2007-2010