Closed nobodxbodon closed 7 years ago
@nobodxbodon What's your goal, to spend as little as possible? :-)
Saving cost would be a side effect of this reviewing, IMO. It's mainly to check (hopefully periodically) whether we make the most out of each cost, by checking the current usage and exploring alternatives. These costs are recuring and can be more easily overlooked overtime.
Besides, we may also consider @kaguillera suggestion of using OS service providers and tip them.
Additionally, it would also save managing cost if any of the service is no longer needed, or if we can use one provider instead of two similar ones.
Cool. I just want to make sure that we're on the same page in terms of the principle of paying for what we use. Gratipay exists to solve the free rider problem in open source, and it is important to me that we model non-free-riding with our own vendors. That puts us at odds with traditional ways of thinking about accounting ... but that's what we're about! :-)
The reasons you mention are valid. I think we should focus on:
Once we have real data and a common way to look at it, then we can circle back here to have the real discussion, about specific financial decisions. Does that sound right to you?
I wonder what you think of questions below:
Besides, I don't quite see dependency on the accounting system or reporting app. Those may have higher priority but this will have immediate influence on the cash flow.
Besides, I don't quite see dependency on the accounting system or reporting app. Those may have higher priority but this will have immediate influence on the cash flow.
Fair enough.
Looking over the list, the ones that jump out at me as providing low value are Pipedrive (#917), Librato (https://github.com/gratipay/gratipay.com/issues/4203), and possibly Papertrail (though I think @rohitpaulk will disagree on that one).
I didn't comment on Pipedrive as marketing is one of our weakness and I hoped the tool would help out. If you think it doesn't quite help with the marketing stategry/methods, +1 with removing it.
Still the items above together is about half of Travis's cost. Now I feel that I must have missed something about Travis. Are we using some features that the commercial version provides, but the free version doesn't? Or is there some special reason that doesn't qualify us from using the free version?
@nobodxbodon We depend heavily on Travis. We need to give back somehow. If we don't give back monetarily, we need to find another way. What would you propose?
As I mentioned above, here are just some ways I can think of. I'm sure there are many other ways.
Those are abstract possibilities, yes. Do you have a concrete proposal for what to replace our monetary payment with? Are you volunteering to do any of the things you mention? :-)
For No.1, how about something like below:
I don't think that's equivalent to $69/mo. Do you?
I'm all for paying for what we use, but this must be value driven.
I don't think $69/mo for Travis is justified at all in the first place - shouldn't it be along the lines of $25 (how much we pay Github)?
Heroku is the one item I definitely wouldn't move - $95/month is definitely worth it. We get database backups, monitoring, easy rollbacks + a reliable service overall. Trying to replicate that on another provider (like AWS) is likely going to cost us as much or more (including man hours).
I don't think that's equivalent to $69/mo. Do you?
IMO it's potentially worth more if compared to ads.
Heroku is the one item I definitely wouldn't move
I agree though there are alternatives to Heroku, I haven't seen any has overall better score, considering feature, cost, maturaity, etc.
I think that paying Travis $70/month is better than dedicating man hours to it that should be dedicated here but I also think that Travis deliberately offers free accounts to open source projects because they have the same value system as we do and that (much like our technical debt discussion) it is useful to have a base goal before you prioritize idealism over bottom line. #imho
FYI, another service we may consider using is Netlify, if necessary, as they also are free specially for OS projects:
they have the same value system as we do
Say more? What I'm seeing is that we have a PWYW pricing model (with a floor), and Travis has a freemium model.
I'm not comfortable being on the "free" side of a freemium business model. Somehow it's related to my even greater discomfort being on the "free" side of an adware business model. I don't know how best to articulate my discomfort right now (I've tried several times today and want to work on other things :). I mean, we've talked about this in years past, so there's almost certainly a record in the old IRC archives if not GitHub. That doesn't mean we don't need to hash it out again with a new Gratipay generation. :-)
Something blah blah, "If you are not paying for it, you're not the customer; you're the product being sold." Blah blah something. But.
I don't think $69/mo for Travis is justified at all in the first place - shouldn't it be along the lines of $25 (how much we pay Github)?
I'd be totally fine switching (paid) CI vendors. The next step there would be a comparison with one or more alternatives. Note that we do use a private project as part of our security workflow.
Redhat offers Resource Grants Program for OS projects to run on their Openshift docker platform. Quoting below:
Projects that would be a great fit for this program
- Research projects for which the results will be made available to the public
- Open Source initiatives for which the code will be made publicly available
- Websites that host useful and publicly available information
- Public service websites
- Web applications which provide free registration and use
- Open Data websites or applications
What's more, they use Jenkins CI
Redhat offers Resource Grants Program for OS projects to run on their Openshift docker platform.
Not really, OpenShift v3, the new version based on Docker, isn't actually launched yet, it's still in developer preview (temporary accounts, can't host real projects on it).
Note that we do use a private project as part of our security workflow.
@whit537 is that a close source project or just a repository to postpone open sourcing the security fixes after they are fixed? If latter it should be still considered OS project I suppose?
Not really, OpenShift v3, the new version based on Docker, isn't actually launched yet, it's still in developer preview
@Changaco thanks I overlooked that. Still the migration from v2 app to v3 seems pretty straightforward, and it can take a while to sign up for the Resource Grants Program. So maybe worthwhile to consider starting the process and putting some part of Gratipay on it.
is that a close source project or just a repository to postpone open sourcing the security fixes after they are fixed? If latter it should be still considered OS project I suppose?
The latter. The question isn't whether we could get free service from Travis, GitHub, etc. We almost certainly could. The question is whether we should as a matter of policy.
@JessaWitzel pretty much captured my thoughts.
What I mean by same value system is that they obviously want to encourage open source work. We are trying to allow people to get paid for open source work and they are trying to support people who are doing open source work by offering them value for free. "Travis CI is, and always will be, free for open source projects." I don't see an ethical dilemma in taking advantage of the value that they want to offer us until we are bootstrapped (am I using that word right?). I do, however, see why you would want to follow the "you should contribute in some way to the open source code that you use" mantra and set a good example. I just think that Travis has intentionally set this pricing structure up because they want us to be able to take advantage of it, which makes it a little different? << The question mark is there because maybe it's not different.
until we are bootstrapped
Alright, so what's our deadline for paying Travis? If we stop paying them now, will we ever start paying them again? When? How will we know?
If we stop paying them now, will we ever start paying them again?
I suppose you mean we'll take the free services for granted and forget about them while using them. If we review services we use periodically (how about twice a year?), that will give good opportunity to decide whether we are in good position to start paying for certain free services.
When? How will we know?
I don't see any other way than to decide case by case. Similarly, it's hard to tell how much is proper to give to a certain free service to show gratitude.
What would we use $69/mo (~$17/wk) for, if it wasn't going to Travis? Would it be added to https://gratipay.com/Gratipay/distributing/? Would it be used for some other vendor? Would it be saved up for conferences/other travel to drum up business?
I think that question needs to be answered in order to determine whether money spent on Travis is worth more than money spent elsewhere.
Would Travis be willing to open a project on Gratipay so open-source projects (like ours) can give them an affordable amount, even though it isn't necessary per Travis' open-source-is-free rule? PWYW for the free part of a Freemium model?
I think that question needs to be answered in order to determine whether money spent on Travis is worth more than money spent elsewhere.
If we're going to accept free services from Travis, it'll be because we're willing to accept free services in general, which means that our cost for GitHub, Sentry, and possibly others (Heroku? Transifex? Librato?) would go away as well.
Would Travis be willing to open a project on Gratipay so open-source projects (like ours) can give them an affordable amount, even though it isn't necessary per Travis' open-source-is-free rule? PWYW for the free part of a Freemium model?
No. I've actually talked to rkh about it, most recently at #534. They're off and running and it would be more hassle than it'd be worth to accept money PWYW.
our cost for GitHub, Sentry, and possibly others (Heroku? Transifex? Librato?) would go away as well.
... unless there's a reason why we would keep paying some vendors that offer free services and not others? I suppose it could be because of case-by-case considerations: this vendors needs it more than this other. I think the way we know whether a vendor needs it is that they stop offering free services, or clearly communicate that they are PWYW. In fact, Read the Docs is a vendor that falls into this latter category. We should probably start paying them!
I suppose it could be because of case-by-case considerations: this vendors needs it more than this other. I think the way we know whether a vendor needs it is that they stop offering free services, or clearly communicate that they are PWYW.
+1
Besides, is the option of listing all free services on bottom of homepage on the table? Something like:
Gratifully powered by: Heroku TravisCI GitHub ...
What would we use $69/mo (~$17/wk) for, if it wasn't going to Travis?
@mattbk IMO it should be part of budget planning, which I don't think there's a formal one for now? Like what we have, how much is spending, how much to spend and how & when, etc.
Anyways, as I mentioned IMO there are many possibilities to spend that in a more direct way to help Gratipay grow.
Besides, is the option of listing all free services on bottom of homepage on the table?
How about listing all services that we use?
How about listing all services that we use?
+1. How about highlighting those that are only free because of our open source nature?
I love all of this. FYI
Any action items for this?
I think we're ready to move forward with applying for free service from various vendors. I've started with Sentry.
Besides, is the option of listing all free services on bottom of homepage on the table?
I don't like listing free services but not paid services, and I'd rather not clutter up the bottom of the homepage if we don't have to. I think the bigger picture here is that we need to develop an internal capacity to be aware of the vendors we're depending on, and make sure we're giving back to them one way or another. Some services may specifically request that we give them a link on our homepage. Let's see what happens when we get into it here ...
I've canceled our Travis subscription, and emailed Nadia at GitHub to ask about free private repos.
We offer Transifex for free to Open Source projects that have no funding, revenue, and/or commercialization model. If you have a question whether your organization qualifies for our Open Source program, please contact us.
https://www.transifex.com/pricing/
Should we contact them?
Should we drop PagerDuty? We already have Uptime Robot—that's where Pager Duty gets its events from. If we drop PagerDuty then we only have email notifications instead of text messages. Can we live with that, @clone1018 @rohitpaulk?
I think it would be stretching to stop paying for Transifex under that definition.
We can buy SMS messaging from Uptime Robot according to this schedule:
n | $ | $ / ea |
---|---|---|
50 | 15 | 0.30 |
100 | 25 | 0.25 |
250 | 55 | 0.22 |
500 | 100 | 0.20 |
I am looking at the logs in Uptime Robot to get a sense of how fast we would use messages (PagerDuty was sending us a lot there for a while, though I think we reconfigured to back off?). If they only send one message per event then 50 messages would last us a while, even if we're sending three messages per event (me, @rohitpaulk, and @clone1018).
Event,Date-Time,Reason,Duration,"Duration (in mins.)"
Up,"2016-11-19 19:14:14",OK,"717 hrs, 53 mins",43074
Down,"2016-11-19 18:40:53","Service Unavailable","0 hrs, 33 mins",33
Up,"2016-09-28 08:45:34",OK,"1257 hrs, 55 mins",75475
Down,"2016-09-28 08:35:44","Connection Timeout","0 hrs, 9 mins",10
Up,"2016-09-28 08:30:58",OK,"0 hrs, 4 mins",5
Down,"2016-09-28 08:29:44","Connection Timeout","0 hrs, 1 mins",1
Up,"2016-09-28 08:24:43",OK,"0 hrs, 5 mins",5
Down,"2016-09-28 08:23:43","Connection Timeout","0 hrs, 1 mins",1
Up,"2016-09-28 08:18:43",OK,"0 hrs, 5 mins",5
Down,"2016-09-28 07:52:53","Connection Timeout","0 hrs, 25 mins",26
Up,"2016-09-28 07:48:02",OK,"0 hrs, 4 mins",5
Down,"2016-09-28 07:47:08","Service Unavailable","0 hrs, 0 mins",1
Up,"2016-09-04 07:54:38",OK,"575 hrs, 52 mins",34553
Down,"2016-09-04 06:17:16","Service Unavailable","1 hrs, 37 mins",97
Up,"2016-07-20 10:20:02",OK,"1099 hrs, 57 mins",65997
Down,"2016-07-20 10:17:42","Connection Timeout","0 hrs, 2 mins",2
Up,"2016-07-07 11:11:47",OK,"311 hrs, 5 mins",18666
We can also use carrier-supplied email-to-SMS for a number of carriers, including Idea Cellular in India, and dozens of options in the USA. Their disclaimer:
Important! Uptime Robot uses free e-mail to SMS gateways of the providers that support it. These gateways are not always stable and such free SMS notifications may not always work as expected.
PagerDuty is pretty expensive for how much we're using it. I propose we turn off PagerDuty and buy a block of 50 SMS messages at Uptime Robot and see how we get along with that. Eh?
I think that's a great idea.
Gosh. I can't find that PagerDuty is even charging us. On their website I find one invoice from 2014, and I don't find anything in PNC in the past couple months. 😳
Last time I got a PagerDuty alert was 7/7/14
I think that was because you were off the rotation, though. @rohitpaulk and I have definitely gotten them more recently.
Grepping our bank logs only turns up that one charge, back in 2014. Sorry, PagerDuty? 😕
Okay! Well, let's give @rohitpaulk a day or three to respond, but I think it looks like we're going to whack PagerDuty and go with SMS directly from Uptime Robot.
We're sorry to see you go! Your subscription has been cancelled. This email is to confirm your subscription with Travis CI has been cancelled and is set to expire on December 29, 2016. If you cancelled by mistake, please respond to this email, and we'll help you getting back up and running. Thank you, The Travis CI Team
I've removed Pipedrive ($25.00) and Travis ($70.93) from the budget, which already takes us from
$127.40/wk = $553.96/mo = $6,647.51/yr
down to
$105.34/wk = $458.03/mo = $5,496.35/yr.
That's a savings of 17.3%!
Let's budget $25/yr for Uptime Robot. Since apparently we haven't been paying PagerDuty for some reason, that's a slight increase. Also, we never adjusted the budget when @gratipay-bot came online, so that's a slight increase from $5/mo to $10/mo. With those adjustments, if Sentry and GitHub are kind to us then we will land at $85.53/wk = $371.91/mo = $4,462.95/yr, or two-thirds of where we were before this exercise.
!m @nobodxbodon *
💃
Hey Gratipay,
We’ve just upgraded your account (Gratipay) to our Sponsored plan. If you have any other questions let us know.
Thanks for choosing Sentry!
p.s. We’d appreciate if you’d tell your friends, or maybe toss up a Sentry link in your app to help us spread the word! You can always find high quality copies of our logo on our Branding page (https://sentry.io/branding/).
What is Paramount
?
Reticket from #878. Below is based on budget spreadsheet, https://github.com/gratipay/inside.gratipay.com/issues/73#issuecomment-58228856, http://inside.gratipay.com/appendices/tools. Please list other missing in it.
As I'm not familiar with many services below, or the past related decisions, any input is more than welcome.