Disclaimer: Matt Asay works for AWS however the views expressed herein are his and do not mirror these of his employer.
In two latest weblog posts (“Placing a steadiness with ‘open’ at Snowflake” and (“Where open helps and where it hurts”), Snowflake spent 6,064 phrases arguing a quite simple idea: All software program needn’t be open—open supply, open requirements, open APIs. It’s not a very objectionable argument and displays the fact that whereas nearly all software program contains open supply code, most software program isn’t licensed as open supply. Snowflake, in different phrases, is safely inside its rights to maintain its software program closed.
And but the corporate clearly felt the necessity (twice) to justify its resolution, reflecting the sturdy gravitational pull of open supply, open requirements, and open APIs, even when its clients don’t seem like clamoring for them.
Open sourcing knowledge
Almost a decade in the past, Cloudera Co-founder Mike Olson made a bold declaration: “No dominant platform-level software program infrastructure has emerged within the final 10 years in closed supply, proprietary kind.” Olson was largely appropriate. Splunk had emerged in that point and maybe just a few different examples, however, on steadiness, he was proper.
Quick ahead to 2021 and Olson’s pronouncement has remained fairly correct with few exceptions. Snowflake is considered one of them. The corporate that payments itself as the info cloud firm has managed to construct a giant enterprise with a proprietary SaaS providing in an trade awash in distinctive open supply knowledge infrastructure like Apache Hadoop, Apache Arrow, Apache Spark, and extra.
This maybe displays a extra nuanced actuality: Enterprises might intuitively need “open” however they place an even bigger premium on “working.” This has been clear for years as corporations have launched managed companies to make it simpler to eat open supply software program or, within the case of corporations like Fauna and Snowflake, present managed companies that aren’t based mostly on open supply in any respect. Getting each “open supply” and “operationally simple” in the identical service is the holy grail, but when enterprises should select one, they’re going to select the answer that’s best for them. In any case, a buyer can flip to Apache Spark, Dremio, or any variety of instruments to construct knowledge warehouses or knowledge lakes, but hundreds of consumers spent roughly half a billion dollars with Snowflake final yr.
So why is Snowflake defending a place that its clients appear to love?
That’s lots of phrases
Between the 2 posts, Snowflake spent lots of effort (three,798 phrases on the Snowflake weblog and a couple of,266 on the InfoWorld put up) to say “We don’t suppose all the pieces must be open.” That’s lots of digital ink spilled to obfuscate a transparent and completely acceptable message that just about each vendor on the planet agrees with. For instance, within the InfoWorld weblog the corporate touts the wonderful contributions its workers have made to the open supply database FoundationDB, which the corporate makes use of in its infrastructure. Nice!
However then it follows that assertion with an ungainly add-on: “Nonetheless, we don’t extrapolate from this to say there’s an inherent benefit to open supply software program.” The authors then double down on the argument that “open isn’t a panacea. We try to keep away from misguided purposes of open that create expensive complexity as an alternative of low-cost ease of use.”
The corporate merely intends (and in the end says) that open supply is a way, not an finish. That’s true! However alongside the way in which it additionally makes errant claims about open supply, suggesting that it by some means would diminish the corporate’s capacity to safe their software program, which merely isn’t true. “At Snowflake, we imagine within the worth of open requirements and open supply, but in addition within the worth of information governance and safety,” the corporate’s co-founder says within the InfoWorld weblog. That “however” is wholly pointless and implies that open requirements and open supply undermine knowledge governance and safety. Neither is true.
There’s additionally the false premise that supply code have to be helpful to all to be helpful in any respect. On the corporate weblog, the authors say, “The question processor of a classy knowledge platform is usually constructed by dozens of PhD program graduates, advanced, refined, and optimized over years. Supply code availability might not considerably enhance the flexibility to understand its inside workings.”
Michael Fischer, a containers professional at AWS, picks up on this: “Open supply was not about enabling customers to grasp and improve the software program. It is about enabling the world to take action. Simply because comparatively few individuals are able to understanding or patching Linux kernel code doesn’t suggest its openness has had little influence. It is slightly smug and insulting to counsel that they should not share as a result of solely PhDs would perceive it. In reality, science advances by way of sharing and publication. That is the entire level of scientific journals and conferences. The artwork advances by way of disclosure.”
Fischer is appropriate, however after all, there’s no regulation stipulating that Snowflake should and even ought to open its code, file codecs, or anything. Dave McCrory, VP of Development and World Head of Insights and Analytics at Digital Realty, and a longtime cloud and open supply observer, points out, “Not all software program must be or must be open sourced. Open supply is an applicable license/mannequin for lots of software program however not all.”
Whether or not Snowflake ought to is in the end a call for its clients, and based mostly on revenues, it appears that evidently Snowflake’s clients don’t care. So once more, why write the posts?
Promoting previous the shut
Most of Snowflake’s big competitors additionally supply proprietary knowledge cloud/platform companies. (Disclosure: I work for AWS, which is a Snowflake companion and competitor, although I’m not concerned with that a part of our enterprise.) It’s extremely unlikely, for instance, that Oracle salespeople are beating up Snowflake for providing proprietary software program. Maybe the stress is coming from Databricks or different open supply distributors?
Databricks recently launched its Delta Sharing venture, an open protocol for securely exchanging giant knowledge units in actual time. This was simply considered one of Databricks’ bulletins on the Data + AI Summit, which sported the tagline, “The long run is open.” Neither is Databricks alone in positioning its knowledge cloud as an open different to options like Snowflake. Journalist Sean Kerner told me, “It is best to see my inbox… Each different pitch is ‘X is an open different to Snowflake.’ ”
Snowflake, for its half, is adamant that open shouldn’t be the right reply in file codecs, supply code, and extra. Not at all times, anyway. Possibly it’s appropriate. However writing hundreds of phrases arguing towards open, versus merely demonstrating worth to clients by way of its choices, is poor advertising and marketing. As I wrote in 2020 concerning the Snowflake IPO:
“Builders have by no means been overly spiritual about open supply. The rationale for [Olson’s comment about a] ‘beautiful’ development is just that open supply made it simpler for builders to get their jobs carried out because of high-quality, simply accessible, open supply knowledge infrastructure. There are, after all, different advantages, such because the communities that always accompany open supply initiatives, coupled with a want to have extra granular management of 1’s software program stack. However in the end open supply has gained as a result of it permits builders to ’get —- carried out.’ Which is why, for instance, you’ll discover builders comfortable to make use of open supply software program like Apache Airflow to load knowledge into their proprietary Snowflake knowledge platform. It’s not cognitive dissonance. It’s pragmatism.”
By rationalizing its choices reasonably than merely delivering worth to clients, Snowflake finally ends up complicated greater than it clarifies. Enterprises clearly admire what it’s promoting. No want for apologies about not being open sufficient.
Copyright © 2021 IDG Communications, Inc.