[RFC] GNU/Weeb Financial Transaction Record Design

[RFC] GNU/Weeb Financial Transaction Record Design

[Alviro Iskandar Setiawan]

Hi,

In order to keep the transaction record consistent, we can use a
chaining hash system that generates the transaction ID by using its
parent ID and the current transaction details.

For example, to create a new transaction, we have several things to note, like:

  1. date
  2. description
  3. commiter
  4. amount

  [ other metadata stuff may be included ]

We can then generate the transaction id by using a hash function,
let's say SHA1:

 trx_info = date . description . commiter . amount . parent_trx_id
 trx_id   = SHA1(info)
 SAVE(trx_id, date, description, commiter, amount, parent_trx_id)

Note that since the current trx ID uses its parent trx ID to generate
itself, it makes the situation impossible to rewrite the history
without breaking the whole chain. That's the point of the consistency
in this design. We can verify that the data is consistent by just
re-hashing the transaction info together with its parent trx id.

Also note that, for the first transaction, it doesn't have a parent
trx id. So only the first transaction doesn't use a parent trx id to
generate itself. Let's call it the genesis transaction.

Please comment!

tq

-- Viro

Re: [RFC] GNU/Weeb Financial Transaction Record Design

[Alviro Iskandar Setiawan]

I haven't started writing the code, but if the idea looks good. I can
start writing the MySQL DDL.

tq

-- Viro

Re: [RFC] GNU/Weeb Financial Transaction Record Design

[Ammar Faizi]

On 6/15/22 4:21 PM, Alviro Iskandar Setiawan wrote:
> Hi,
> 
> In order to keep the transaction record consistent, we can use a
> chaining hash system that generates the transaction ID by using its
> parent ID and the current transaction details.
> 
> For example, to create a new transaction, we have several things to note, like:
> 
>    1. date
>    2. description
>    3. commiter
>    4. amount
> 
>    [ other metadata stuff may be included ]
> 
> We can then generate the transaction id by using a hash function,
> let's say SHA1:
> 
>   trx_info = date . description . commiter . amount . parent_trx_id
>   trx_id   = SHA1(info)
>   SAVE(trx_id, date, description, commiter, amount, parent_trx_id)
> 
> Note that since the current trx ID uses its parent trx ID to generate
> itself, it makes the situation impossible to rewrite the history
> without breaking the whole chain. That's the point of the consistency
> in this design. We can verify that the data is consistent by just
> re-hashing the transaction info together with its parent trx id.
> 
> Also note that, for the first transaction, it doesn't have a parent
> trx id. So only the first transaction doesn't use a parent trx id to
> generate itself. Let's call it the genesis transaction.
> 
> Please comment!

Hi Al,

I like the idea of chaining the transaction ID to make the history
consistent and immutable. Overall it looks good to me, let's discuss
this further.

Several questions below:

  1) What are the exact fields we are going to save?

  2) How do we maintain a new field in the future if we ever add more?

  3) How do we distribute the trusted chains to public?

For no (2), you must not break the old records that don't have the new
field.

For no (3), I think it's quite simple, just post the transaction
history record periodically to the group. We can use a Telegram bot
for that.

Comment?

[ Anyway, you can start writing the code. ]

-- 
Ammar Faizi

Re: [RFC] GNU/Weeb Financial Transaction Record Design

[Alviro Iskandar Setiawan]

On Wed, Jun 15, 2022 at 4:42 PM Ammar Faizi <ammarfaizi2@gnuweeb.org> wrote:
> Several questions below:
>
>   1) What are the exact fields we are going to save?

I will post the DDL with exact fields we are going to save to answer
this question.

>   2) How do we maintain a new field in the future if we ever add more?

Maybe just make them be an empty string. Appending an empty string
won't change the hash because it doesn't change anything. For example,
we create a new field x, make it nullable, if it's NULL then when
we're validating the old records that don't have this field just treat
it as an empty string, it won't change the hash, so it's backward
compatible.

>   3) How do we distribute the trusted chains to public?
>
> For no (2), you must not break the old records that don't have the new
> field.
>
> For no (3), I think it's quite simple, just post the transaction
> history record periodically to the group. We can use a Telegram bot
> for that.

I agree with you.

> [ Anyway, you can start writing the code. ]

oc

tq

-- Viro

Re: [RFC] GNU/Weeb Financial Transaction Record Design

[Ammar Faizi]

On 6/15/22 5:00 PM, Alviro Iskandar Setiawan wrote:
>>    2) How do we maintain a new field in the future if we ever add more?
> 
> Maybe just make them be an empty string. Appending an empty string
> won't change the hash because it doesn't change anything. For example,
> we create a new field x, make it nullable, if it's NULL then when
> we're validating the old records that don't have this field just treat
> it as an empty string, it won't change the hash, so it's backward
> compatible.

Not sure how will it look like. I am looking forward to the patchset then.

-- 
Ammar Faizi

Re: [RFC] GNU/Weeb Financial Transaction Record Design

[Alviro Iskandar Setiawan]

On Wed, Jun 15, 2022 at 5:03 PM Ammar Faizi <ammarfaizi2@gnuweeb.org> wrote:
> On 6/15/22 5:00 PM, Alviro Iskandar Setiawan wrote:
> >>    2) How do we maintain a new field in the future if we ever add more?
> >
> > Maybe just make them be an empty string. Appending an empty string
> > won't change the hash because it doesn't change anything. For example,
> > we create a new field x, make it nullable, if it's NULL then when
> > we're validating the old records that don't have this field just treat
> > it as an empty string, it won't change the hash, so it's backward
> > compatible.
>
> Not sure how will it look like. I am looking forward to the patchset then.

Yeeeeeeh, just sit and wait for my patchset, you can comment later on :p

tq

-- Viro