3 ms·
A few of these make unwarranted assumptions, though this may be based on the kinds of companies they invest in. Examples: S1.3b - For several types of software
by jandrewrogers 7y ago
A few of these make unwarranted assumptions, though this may be based on the kinds of companies they invest in. Examples:
S1.3b - For several types of software, releasing continuously is an anti-pattern. Not only will it annoy your customers, it will significantly increase the cost of development. Not every tech startup is a website.
S1.4b - Same as above, not every kind of software application can be properly monitored without custom monitoring development, though most can. If someone develops a custom monitoring solution, you should at least understand why first; using an off-the-shelf solution in some cases would be a negative.
S2.6g - There should be an on-premise option. For several industries and applications it is required by the customers and/or IaaS has substantial adverse economics. The correct answer is that there is a financial model that captures these CapEx/OpEx tradeoffs. I've seen too many startups burn millions of dollars per year unnecessarily because they never did the math. (You should always be able to easily deploy across IaaS and on-prem though.)
- momokoko 7y agoIts more that they don't track "releases". In my experience, early companies (that are not hardware related) that track full releases are very top heavy with non technical leadership. This is because technical people generally understand in modern saas, releases aren't really an actual thing with early stage companies as they are constantly patching bugs and small features. The point about monitoring is more a "bus factor". If the organization in general has good visibility into the operation of the software, there tends to be less situations where only one person knows how a specific recurring issue gets resolved.
- jandrewrogers 7y agoThere are many cases where the customer requirements dictate release adoption timelines (for business, regulatory, and/or logistical reasons), even on SaaS platforms. The way you deal with this in practice is running several releases in parallel in your production environment. This is messy and expensive so you optimize your release process to minimize the number of versions you have to support. The other common scenario is when deploying new releases is intrinsically extremely expensive in one or more dimensions. Some data infrastructure SaaS can be like this. Again, in these situations startups correctly optimize their release model differently than e.g. a website.
- mason55 7y agoRight, if you’re a high touch enterprise SaaS that requires things like customer training, good luck with pure CD. Your customers are going to be very cranky if stuff is changing all the time. That’s not to say you can’t get many of the good parts of CD but you’re going to be making releases and supporting a few different versions at all times. There’s just no way around it.
- Cthulhu_ 7y agoRe: continuous release, you can also replace the term with continuous integration; not so much if you DO release regularly, but if you CAN. Ask questions like, do we have build monitoring on the release branch? Is it green most of the time? Is work integrated into it regularly (multiple times a day)? If we discover a bug, can we just push a fix to master and deploy or do we need to do a hotfix on an older tag? What are the steps to create a new release? While yes, continuous deployment may not be desirable in a lot of cases, continuous integration is a practice that (IMO) most software should strive for.